Disease data analysis method and lung cancer risk prediction system based on medical knowledge base

Abstract

The present invention discloses a disease data analysis method based on a medical knowledge base and a lung cancer risk prediction system, the method comprising: obtaining a disease big data cohort from a distributed medical information database according to a medical dictionary; standardizing the disease big data cohort according to a disease classification standard; establishing a lung cancer disease cohort based on the disease big data cohort; and screening risk indicators related to lung cancer based on correlation analysis according to the lung cancer disease cohort. A risk prediction model can be constructed based on the risk indicators. Based on a medical knowledge base, the present invention realizes data extraction and structuring of disease data and lung cancer-related data, providing data guarantee for subsequent lung cancer risk prediction.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical big data processing, and in particular relates to a disease data analysis method and a lung cancer risk prediction system based on a medical knowledge base. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] Lung cancer occurs in the bronchial mucosal epithelium and is also called bronchial cancer. In the past 50 years, many reports have shown that the incidence of lung cancer has increased significantly. Among male cancer patients, lung cancer ranks first, and the incidence in women is increasing rapidly, ranking second or third among common malignant tumors in women.

[0004] At present, the auxiliary diagnosis methods related to lung cancer mainly focus on image analysis based on deep learning and influence index analysis based on traditional statistical analysis. Among them, the application of deep learning technology to early lung cancer detection mainly includes data preprocessing and the use of convolutional networks to detect lung nodules. However, these methods mainly focus on identifying lung nodule areas and non-lung nodule areas. On the one hand, the types and probabilities of lung nodule canceration are closely related to the characteristics of lung nodules and other related disease variables. Traditional detection convolutional networks can only derive the total probability of patients suffering from lung cancer, and cannot distinguish the related disease variables of lung cancer. Doctors are often required to evaluate. Traditional models are mainly based on epidemiological survey data, such as age, smoking status and occupation. Information bias is inevitable in the process of information collection; and although there are many sources of medical data, the data forms are diverse. Each region has multiple hospitals, clinics and other medical institutions, and each medical institution has its own habits for recording diseases, resulting in messy data, large workload for screening and sorting data related to lung cancer, and easy omissions, and lack of effective data normalization methods. Summary of the invention

[0005] In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a disease data analysis method and a lung cancer risk prediction system based on a medical knowledge base, which realizes the data extraction and structuring of disease data and lung cancer related data, and provides data guarantee for subsequent lung cancer risk prediction.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] A disease data analysis method based on a medical knowledge base comprises the following steps:

[0008] Obtain disease big data cohorts from distributed medical information databases based on medical dictionaries;

[0009] According to the disease classification standards, the data of the disease big data cohort is standardized;

[0010] Establish a lung cancer disease cohort based on the disease big data cohort;

[0011] According to the lung cancer disease cohort, risk indicators associated with lung cancer were screened based on correlation analysis.

[0012] One or more embodiments provide a lung cancer risk prediction system, including a cloud platform and a distributed database system, wherein the cloud platform includes:

[0013] The disease cohort acquisition module obtains disease big data cohorts from distributed medical information databases based on medical dictionaries;

[0014] The data standardization module standardizes the disease big data cohort according to the disease classification standards;

[0015] The disease cohort establishment module establishes a lung cancer disease cohort based on the disease big data cohort;

[0016] The risk indicator analysis module screens risk indicators related to lung cancer based on correlation analysis according to the lung cancer disease cohort; the risk indicators screened include chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, and asthma;

[0017] The risk prediction model building module builds a lung cancer risk prediction model based on the screened risk indicators;

[0018] The lung cancer probability prediction module receives the risk prediction request, retrieves the relevant historical disease data queue, and obtains the lung cancer incidence probability prediction result based on the lung cancer prediction model.

[0019] One or more of the above technical solutions have the following beneficial effects:

[0020] The present invention targets medical big data with complex sources, extracts disease data and lung cancer disease data based on a medical knowledge base, and realizes structuring and standardization, providing big data support for subsequent analysis.

[0021] The present invention is also based on a disease big data cohort. With the support of disease big data, it fully explores the risk indicators related to lung cancer, tries to ensure the comprehensiveness of the risk indicators, and largely makes up for the subjectivity of manual screening alone; and ensures the universality of subsequent prediction models. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 is a flow chart of a disease data analysis method based on a medical knowledge base in an embodiment of the present invention;

[0024] Figure 2 Flow chart of the data standardization method in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0026] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0027] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0028] Embodiment 1

[0029] This embodiment discloses a disease data analysis method based on a medical knowledge base, comprising the following steps:

[0030] Step 1: Obtain disease big data queue from distributed medical information database according to medical dictionary;

[0031] The distributed database system includes medical information databases deployed in various cities. In this embodiment, the medical information database includes a population information database, a public health database, an electronic medical record database, a medical insurance database, a health examination database, a cause of death database, etc. distributed in various cities in Shandong Province.

[0032] Among them, the population information database includes: residents' basic personal information, social security information, housing information and residents' integrity and dishonesty information.

[0033] The public health database includes: basic personal health information, disability form, health examination form, student physical examination form, birth medical certificate, newborn home visit information, child health examination information, prenatal follow-up service information, delivery record form, postpartum visit service information, pregnancy examination record form, vaccination card information, infectious disease report card, occupational disease report card, foodborne disease card, death medical certificate, hypertension patient follow-up form, type 2 diabetes patient follow-up form, severe mental illness patient management form, severe mental illness patient onset information, severe mental illness patient discharge information, coronary heart disease patient information, stroke patient information, tumor patient information, tuberculosis patient follow-up form, childbearing age couple information, marriage information, birth registration and birth approval information, pregnancy information, women's birth history information, contraceptive information, family planning operation information, outflow population information, inflow population information, outpatient summary information, inpatient summary information.

[0034] The electronic medical record database includes: outpatient / emergency registration, outpatient / emergency treatment records, outpatient / emergency / hospitalization western medicine prescriptions, outpatient / emergency / hospitalization Chinese medicine prescriptions, test records, test result details, examination records, admission records, hospitalization medical record homepage, traditional Chinese medicine hospitalization medical record homepage, and discharge records.

[0035] The health examination database includes: a physical examination report, a physical examination item list, and a physical examination item detailed list.

[0036] The medical insurance database includes: basic information, diagnosis and treatment information, total medical insurance expense information, and detailed medical insurance expense information.

[0037] The medical knowledge base includes medical information such as medical dictionaries and disease classification standards of various versions.

[0038] The step 1 specifically includes:

[0039] Step 1.1: According to the medical dictionary in the medical knowledge base, search for a data table in the database system that contains disease-related fields;

[0040] Step 1.2: Based on the found data table, extract fields such as ID number, disease, disease code, onset time, etc., and record the data source of the disease, such as the source city, source data table, ID in the data table, etc., to generate a disease big data queue.

[0041] Step 2: Standardize the disease big data cohort data according to the disease classification standards;

[0042] The specific steps include:

[0043] Step 2.1: Screen the sample data set from the disease big data cohort, compare the disease names in the sample data with the disease names in the disease classification standard, and standardize the disease names in the sample data;

[0044] The step of standardizing the disease names in the sample data includes: creating a standardized name field, and performing standardization in the following steps:

[0045] (1) Name-identical comparison: Obtain sample data whose disease names are completely consistent with the disease names in the disease classification standards, and write the original disease names into the standardized name field.

[0046] (2) Name similarity comparison: obtain sample data whose disease names have a similarity with the disease names in the disease classification standard that exceeds a set threshold, and write the original disease name into the standardized name field; the similarity measurement can adopt existing text similarity methods such as cosine similarity and Euclidean distance, which are not limited here.

[0047] (3) Inclusion control: Obtain sample data in which the disease name has an inclusion relationship with the disease name in the disease classification standard, such as "prostatitis (non-surgical treatment)" and "prostatitis", and write the original disease name into the standardized name field.

[0048] (4) The standardized names of the sample data are manually reviewed by the user through the client. Specifically, during the manual review, the disease names can be sorted by frequency, and disease names with high frequency can be reviewed first.

[0049] During the standardization process, the system automatically records the control methods corresponding to the names of each disease in the disease big data queue. In this embodiment, the sample data set selects medical insurance data with good data quality (about 60,000 items), and the disease classification standard adopts ICD10 coding.

[0050] Step 2.2: For the unstandardized data in the disease big data cohort, compare the disease names with the original disease names in the sample data to complete the standardization of some disease names;

[0051] Specifically, for data whose disease names are identical to the original disease names in the sample data, whose name similarity is greater than a set threshold, or whose disease names are included in the sample data, the standardized name corresponding to the original disease name in the sample data is written into the standardized field.

[0052] Step 2.3: For the remaining unstandardized data in the disease big data queue, compare the disease codes with the codes in the disease classification standards. For the data with successful coding comparison, write the disease name corresponding to the code in the disease classification standards into the standardized field.

[0053] Specifically, the comparison of the disease code with the code in the disease classification standard is carried out in stages: first, it is compared with all 6 digits of the code in the disease classification standard, then it is compared with the first 4 digits, and finally it is compared with the first 2 digits.

[0054] Step 2.4: The standardized names in the disease big data queue are manually reviewed by the user through the client. Due to the large amount of data to be standardized (about 7 million), the disease names can be sorted by frequency here, and only the disease names with larger frequency are reviewed;

[0055] Step 2.5: Count the control rate. If the control rate exceeds the set threshold, standardization ends.

[0056] For medical big data with complex sources, we obtain standardized data of sample data based on multi-level text matching methods, and then complete the standardization of massive data based on the standardized data of the samples in sequence by matching the names and codes. Compared with the direct matching of all medical big data with standard data, we can obtain higher standardization rate and accuracy while taking into account the efficiency of standardization.

[0057] Step 3: Establish a lung cancer disease cohort based on the disease big data cohort.

[0058] The specific steps include:

[0059] is configured to perform the following steps:

[0060] Step 3.1: Retrieve disease names related to lung cancer from the disease big data queue; since there are many forms of expression related to lung cancer, synonym expansion is required here. Those skilled in the art can understand that retrieval can also be performed by constructing logical expressions;

[0061] Step 3.2: The user reviews the retrieved lung cancer-related disease names through the client. Those skilled in the art will appreciate that the review may be performed to delete data records individually or in batches by constructing a logical expression.

[0062] Step 3.3: According to the names of lung cancer-related diseases, match the ID number, gender, region and other data from the disease big data cohort to obtain the lung cancer disease cohort.

[0063] Receive case inclusion criteria and control group matching rules, and obtain lung cancer case and control group data from the lung cancer disease cohort. Obtain lung cancer cases based on the case inclusion criteria, and match the control group for case samples based on the control group matching rules. In this embodiment, the case inclusion criteria are: the first diagnosis of lung cancer is between January 1, 2012 and December 31, 2016, at least one longitudinal observation record, age greater than 20 years old and less than 90 years old, and no cancer at baseline; for all cases, according to the 1:4 ratio principle, match the corresponding controls according to age and gender, use the matched data for data modeling, randomly select 2 / 3 of the cases as the modeling cohort cases, and the remaining 1 / 3 of the cases as the verification cohort cases.

[0064] Step 4: Based on the lung cancer disease cohort, risk indicators related to lung cancer were screened based on correlation analysis.

[0065] The specific steps include:

[0066] Step 4.1: Based on the lung cancer disease cohort, extract disease variables corresponding to lung cancer events;

[0067] The lung cancer-related disease variables include chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, asthma, hypertension, stroke, upper respiratory tract infection, and obesity; heart disease includes coronary heart disease, angina pectoris, myocardial infarction, and heart disease.

[0068] Lung cancer events included: small cell lung cancer, squamous cell carcinoma, adenocarcinoma, bronchial tumors, and lung cancer.

[0069] Step 4.2: Perform correlation analysis on each disease variable and the lung cancer event, and use the disease variables with correlation greater than a set threshold as candidate risk indicators; this embodiment adopts a non-conditional logistic regression model.

[0070] (1) Based on whether or not there is a disease variable, a binary risk indicator matrix X is constructed, where each row corresponds to a person and each column corresponds to a type of risk indicator. The mth row and nth column of the matrix X (m, n) indicates whether the mth person has the nth disease variable. If yes, it is recorded as 1, and if no, it is recorded as 0.

[0071] (2) Based on whether a lung cancer event occurs, a binary lung cancer matrix Y is constructed, where the matrix Y contains one column and each row corresponds to whether a person has a lung cancer event;

[0072] (3) Perform correlation analysis on each column of the binary risk indicator matrix X and the matrix Y to obtain the correlation matrix R. Each element in the matrix R represents the correlation between each disease variable and lung cancer. Disease variables with correlation greater than the set threshold are selected as candidate risk indicators.

[0073] Step 4.3: Based on the Bayesian network, select the final risk indicator from the candidate risk indicators.

[0074] A Bayesian network is a graphical model that represents the probability of connection between variables and can be used to discover potential relationships between data. The result of Bayesian learning is represented as the probability distribution of random variables, which can be interpreted as the degree of trust in different possibilities. In this embodiment, the candidate risk indicators obtained in step 5.1 and lung cancer events are input into the Bayesian network to obtain candidate risk indicators associated with lung cancer events as the final risk indicators.

[0075] Those skilled in the art will appreciate that they can also manually perform auxiliary index screening based on literature, clinical data and national standards, and adopt multiple index screening methods to prevent the omission of important indicators.

[0076] The risk indicators finally selected in this embodiment include: chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, and asthma.

[0077] This embodiment is based on a disease big data cohort and uses univariate correlation analysis, Bayesian network and manual query methods to fully explore the risk indicators related to lung cancer, which largely makes up for the subjectivity of manual screening alone; and, with the support of disease big data, it ensures that risk indicators are not missed and the universality of subsequent prediction models is guaranteed.

[0078] Embodiment 2

[0079] This embodiment discloses a lung cancer risk prediction system. It includes: a distributed database system, a cloud platform, a work terminal and a user terminal. Among them,

[0080] The cloud platform includes:

[0081] The lung cancer incidence probability prediction subsystem includes:

[0082] The disease big data queue acquisition module retrieves the disease big data queue from the distributed database system.

[0083] Data standardization module: standardize the data of disease big data cohort.

[0084] The lung cancer disease cohort acquisition module establishes a lung cancer disease cohort based on the disease big data cohort.

[0085] The risk indicator screening module counts and screens relevant risk indicators based on lung cancer-related disease variables.

[0086] For the specific implementation methods of the above modules, please refer to Example 1.

[0087] The lung cancer risk prediction model building module builds a lung cancer risk prediction model based on the screened risk indicators. It is configured to perform the following steps:

[0088] Step (1) Based on the risk indexes screened, a logistic regression model was used to perform univariate analysis, and independent predictors of lung cancer were selected by stepwise screening. The test level α = 0.05.

[0089] The formula for the logistic regression model is as follows:

[0090]

[0091] Among them, β0 is a constant term, β1, β2, …, β p are regression coefficients, X1, X2, …, X p is the independent variable and P is the predicted value.

[0092] After univariate regression analysis, the screened variables included chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, asthma, hypertension, stroke, upper respiratory tract infection, and obesity; heart disease included coronary heart disease, angina pectoris, myocardial infarction, and heart disease.

[0093] Step (2) conducts multivariate logistic regression analysis on the risk indicators and combines them with the Gail model to establish a lung cancer disease prediction model.

[0094] The Gail model is a mathematical model for calculating the risk of lung cancer based on the risk of lung cancer, the risk of competing events, and the results of a multifactor unconditional logistic regression model in a big data cohort of the entire population of Shandong throughout their life cycle. It converts the relative risk value of lung cancer in an individual into an absolute risk value.

[0095] The formula for Gail's model is as follows:

[0096]

[0097] in is the age-specific incidence of lung cancer, F(t) = 1-AR, AR is the population attributable risk. r(t) is the relative risk, is the competing risk probability of surviving to age t.

[0098] The results of multivariate logistic regression analysis revealed risk factors for chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, and asthma.

[0099] This embodiment constructs the model multiple times, each time introducing a new risk indicator, and measures the prediction performance of the model through the Net Reclassification Index (NRI) to obtain the final prediction model with the best prediction performance. Specifically, firstly, single-factor modeling is performed based on each risk indicator to obtain the initial prediction model with the best prediction performance, and the corresponding risk indicator is the most important factor; then, based on the initial prediction model, one of the other risk indicators is introduced to perform two-factor modeling to obtain the two-factor prediction model with the best prediction performance, and the newly introduced risk indicator is the second most important factor; and so on, new risk indicators are introduced in turn until the performance of the prediction model is no longer enhanced.

[0100] Each time a prediction model is constructed, ROC, sensitivity, and specificity are calculated; then NRI = (sensitivity test2 + specificity test2) - (sensitivity test1 + specificity test1) is calculated as a measure of model performance. If NRI>0, it indicates that after adding new predictors, the prediction ability of the new model has improved, and the correct classification ratio has increased by NRI percentage points. The more NRI increases, the better the variable prediction effect is, and the more important the variable is.

[0101] The model of the present invention is constructed by introducing one risk indicator at a time, gradually determining the risk indicator most relevant to lung cancer, and ensuring the accuracy of prediction.

[0102] Based on this, the prediction model with the best performance was obtained, and the importance of the risk indicators screened was ranked.

[0103] In this embodiment, all statistical analyses were performed using R3.5 language, the statistical description of numerical variables was in the form of mean ± standard deviation, the categorical variables were in the form of frequency (percentage), and the inter-group comparisons were performed using t-test (continuous variables) or chi-square test (categorical variables).

[0104] The area under the ROC curves (AUC) was used to measure the discriminative ability of the model. The R software package "survival" was used to perform Cox risk regression, and the package "pROC" was used to draw the ROC curve and calculate the AUC. Two-thirds of the cohort population was randomly selected as the training set, and the remaining 1 / 3 was used as the validation set for verification.

[0105] Health report generation subsystem, including:

[0106] User management module, used to manage the identity information of registered users;

[0107] Disease response strategy management module, used to store precautions and response suggestions for various diseases;

[0108] A lung cancer probability prediction module receives a prediction request sent by a user terminal, retrieves a historical disease data queue of the user, and obtains a lung cancer incidence probability prediction result based on a lung cancer prediction model;

[0109] Specifically, for each risk indicator variable in the prediction model, if the user suffers from the disease corresponding to the risk indicator, the risk indicator variable is assigned a value of 1, otherwise it is assigned a value of 0, and the probability of lung cancer of the user is calculated.

[0110] The lung cancer risk index analysis module obtains the user's lung cancer risk index and the contribution rate of each risk index;

[0111] Specifically, the contribution rate of each risk indicator is calculated as follows:

[0112] For each risk indicator variable assigned a value of 1, a value of 0 is assigned and the probability of lung cancer is calculated to obtain the probability of lung cancer when the user does not suffer from the disease corresponding to the risk indicator; the probability of lung cancer is subtracted from the probability of lung cancer obtained by the lung cancer probability prediction module to obtain the contribution rate of the disease corresponding to each risk indicator to the user's lung cancer.

[0113] The lung cancer risk indicator guidance module obtains corresponding coping strategies for the diseases that the user suffers from that may affect lung cancer;

[0114] The health report generation module is used to generate a visual report based on health information, lung cancer incidence probability prediction results and lung cancer risk indicator guidance results.

[0115] The cloud platform pre-packages relevant data processing methods. The above data processing is all performed on the cloud platform, and the data will not be transmitted to other terminals, ensuring data security and protecting user privacy.

[0116] The present invention adopts the cloud platform as the core of data aggregation and data processing, and connects with the databases of medical institutions at all levels in cities and counties to ensure the authenticity and integrity of the data, as well as the security of the data.

[0117] The present invention provides a health assessment system for users, which can predict the user's probability of developing lung cancer and the contribution rate of the user's lung cancer-related diseases, provide coping strategies for these diseases, and play a role in guiding users to prevent lung cancer.

[0118] Work terminal, including:

[0119] Data standardization module, used to review the standardization results of sample data and all data in the cloud platform;

[0120] A lung cancer-related disease name acquisition module is used to receive a lung cancer-related disease name input by a user, or a logical expression for retrieving a disease name; and to review the retrieved disease name;

[0121] The risk indicator determination module is used to obtain candidate risk indicators and their Bayesian network structure diagrams from the cloud platform, receive the user's confirmation and correction of the risk indicators, and send them to the cloud platform;

[0122] The model building module receives the case inclusion criteria, control group matching rules, and the model used;

[0123] The model correction module is used to correct the adopted model and model parameters.

[0124] The user terminal includes:

[0125] Login authentication module, used to authenticate user identity;

[0126] The health report viewing module is used to obtain the user's health information from the cloud platform, including historical physical examination information, case information, etc.;

[0127] Lung cancer probability prediction module, used to obtain lung cancer incidence probability prediction results from the cloud platform;

[0128] The lung cancer risk index guidance module is used to obtain the user's lung cancer risk index and the contribution rate of each risk index from the cloud platform;

[0129] The health report generation module is used to generate a visual report based on health information, lung cancer incidence probability prediction results and lung cancer risk indicator guidance results.

[0130] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0132] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A disease data analysis method based on a medical knowledge base, characterized in that: The following steps are involved: According to the medical dictionary, a disease big data queue is obtained from a distributed medical information database; specifically, the following steps are performed: according to the disease words in the medical dictionary, a data table in the database system is searched for a field containing these disease words; based on the found data table, user identity information, disease information, and data source are extracted to generate a disease big data queue; According to the disease classification standards, the disease big data cohort is standardized; specifically, the sample data set is screened from the disease big data cohort, the disease names in the sample data are compared with the disease names in the disease classification standards, and the disease names in the sample data are standardized; For the unstandardized data in the disease big data queue, the disease name is compared with the original disease name in the sample data. For the data with successful comparison, the corresponding completed part of the standardized disease name in the sample data is written into the standardized field; For the remaining unstandardized data in the disease big data queue, the disease codes are compared with the codes in the disease classification standards. For the data with successful code comparison, the disease names corresponding to the codes in the disease classification standards are written into the standardized fields; Based on the disease big data cohort, a lung cancer disease cohort was established; the specific steps are: Retrieve disease names related to lung cancer from the disease big data cohort; due to the large number of lung cancer-related expressions, synonym expansion is required; The user reviews the retrieved lung cancer-related disease names through the client; According to the names of lung cancer-related diseases, the ID number, gender, and regional data were matched from the disease big data cohort to obtain the lung cancer disease cohort; the case inclusion criteria and control group matching rules were received to obtain lung cancer case and control group data from the lung cancer disease cohort; lung cancer cases were obtained based on the case inclusion criteria, and the control group was matched for the case samples based on the control group matching rules; Based on the lung cancer disease cohort, risk indicators related to lung cancer are screened based on correlation analysis; specifically, based on the lung cancer disease cohort, disease variables corresponding to lung cancer events are extracted; The correlation between each disease variable and lung cancer outcome events was analyzed, and the disease variables with correlation greater than the set threshold were used as candidate risk indicators; Based on the Bayesian network, the risk indicators for modeling are screened from the candidate risk indicators; The obtained candidate risk indicators and lung cancer events are input into the Bayesian network to obtain the candidate risk indicators associated with lung cancer events as the final risk indicators, and multiple indicator screening methods are used to prevent the omission of important indicators. The risk indicators most relevant to lung cancer are gradually determined by introducing one risk indicator at a time. The risk indicators finally selected included: chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, and asthma; Based on a big disease data cohort, univariate correlation analysis, Bayesian network and manual query methods were used to fully explore the risk indicators associated with lung cancer.

2. A disease data analysis method based on a medical knowledge base as claimed in claim 1, characterized in that: The standardization of the disease names in the sample data includes: (1) Name-identical comparison: Obtain sample data with disease names that are completely consistent with those in the disease classification standard, and write the original disease name into the standardized name field; (2) Name similarity comparison: obtain sample data whose disease names are similar to disease names in the disease classification standard and exceed the set threshold, and write the original disease names into the standardized name field; (3) Inclusion control: obtaining sample data in which the disease names have an inclusion relationship with the disease names in the disease classification standards; (4) The standardized names of the sample data are manually reviewed by the user.

3. A lung cancer risk prediction system, characterized in that: Based on the method as claimed in claim 1 or 2, including a cloud platform and a distributed database system, the cloud platform includes: The disease queue acquisition module acquires the disease big data queue from the distributed medical information database according to the medical dictionary; specifically, it includes: searching for data tables in the database system that have fields containing these disease words according to the disease words in the medical dictionary; extracting user identity information, disease information, and data source based on the found data table to generate the disease big data queue; The data standardization module standardizes the disease big data cohort according to the disease classification standard; specifically, it includes: screening the sample data set from the disease big data cohort, comparing the disease names in the sample data with the disease names in the disease classification standard, and standardizing the disease names in the sample data; For the unstandardized data in the disease big data queue, the disease name is compared with the original disease name in the sample data. For the data with successful comparison, the corresponding completed part of the standardized disease name in the sample data is written into the standardized field; For the remaining unstandardized data in the disease big data queue, the disease codes are compared with the codes in the disease classification standards. For the data with successful code comparison, the disease names corresponding to the codes in the disease classification standards are written into the standardized fields; The disease cohort establishment module establishes a lung cancer disease cohort based on the disease big data cohort; The risk indicator analysis module screens risk indicators related to lung cancer based on correlation analysis according to the lung cancer disease cohort; specifically, it extracts disease variables corresponding to lung cancer events based on the lung cancer disease cohort; The correlation between each disease variable and lung cancer outcome events was analyzed, and the disease variables with correlation greater than the set threshold were used as candidate risk indicators; Based on the Bayesian network, the risk indicators for modeling are screened from the candidate risk indicators; The risk indicators screened included chronic obstructive pulmonary disease, heart disease, diabetes, rhinitis, emphysema, chronic kidney disease, chronic bronchitis, pneumonia, and asthma; The risk prediction model building module builds a lung cancer risk prediction model based on the screened risk indicators; The lung cancer probability prediction module receives the risk prediction request, retrieves the relevant historical disease data queue, and obtains the lung cancer incidence probability prediction result based on the lung cancer prediction model.

4. A lung cancer risk prediction system according to claim 3, characterized in that: The calculation method of the lung cancer incidence probability prediction result is: For each risk indicator variable in the prediction model, if the patient suffers from the disease corresponding to the risk indicator, the risk indicator variable is assigned a value of 1, otherwise it is assigned a value of 0, and the probability of lung cancer is calculated.

5. A lung cancer risk prediction system according to claim 4, characterized in that: The lung cancer probability prediction module also includes obtaining risk indicators of lung cancer events and the contribution rate of each risk indicator, and generating a health report.

6. A lung cancer risk prediction system according to claim 5, characterized in that: The system also includes a user terminal connected to the cloud platform, which is used to generate an incidence risk prediction request, and receive and visualize the lung cancer incidence probability prediction results sent by the cloud platform.

Images