Method and system for detecting fatty liver patient missed in liver color ultrasound examination

Abstract

The invention discloses a method and a system for detecting fatty liver patients missed in liver color ultrasound examination. The method comprises the following steps: collecting sample data of physical examinees including liver color ultrasound examination, FibroTouch detection, biochemical detection, physique measurement and questionnaire survey data; carrying out data cleaning on the obtained sample data, wherein the data cleaning comprises deletion of missing values and elimination of non-target physical examination data; establishing a preliminary binary logistic regression model based on full variables; in combination with various methods including a logistic regression back-off method, a stepwise regression method and a ridge regression method, screening variables most valuable to the model from the full variables; variable coding is carried out, and the fatty liver based on FibroTouch diagnosis is used as a result variable to establish a diagnosis model; the diagnosis model recognizes a person who is not diagnosed as a fatty liver in the liver color ultrasound examination, and outputs whether the person is a non-alcoholic fatty liver or not as a diagnosis result.

Description

Technical Field

[0001] The present invention belongs to the field of medical health examination, and more specifically, relates to a data processing method, a data processing device, a data processing equipment and a computer-readable storage medium for diagnosing patients with fatty liver who are missed by liver color ultrasound examination during health examination. Background Art

[0002] With the prevalence of obesity and type 2 diabetes, non-alcoholic fatty liver disease (NAFLD) has become the most common liver disease in the world, affecting about 37.8% of the world's population. In China, this proportion is 44.4%. NAFLD includes a series of diseases, from simple hepatic steatosis to non-alcoholic steatohepatitis, the latter of which can develop into liver fibrosis, cirrhosis, and even liver cancer. At the same time, NAFLD can promote the occurrence and development of cardiovascular and cerebrovascular diseases and chronic kidney disease. Therefore, NAFLD has become an increasingly serious public health problem in my country.

[0003] Nearly 500 million people in China complete health checkups every year, but the current liver color Doppler ultrasound examination as the main screening method has obvious limitations. The sensitivity of liver color Doppler ultrasound to fatty liver, especially mild fatty liver, is limited, which can easily lead to missed diagnosis. This is due to the characteristics of the color Doppler ultrasound imaging principle itself. For cases where the degree of fatty degeneration of liver cells is mild (≥5% can be defined as fatty liver), it may not be accurately identified. Although liver color Doppler ultrasound is widely used in health examinations, this missed diagnosis may lead to a large number of fatty liver patients not being discovered and diagnosed in time, thereby delaying the treatment. Furthermore, this may cause the patient's condition to gradually progress, increasing the risk of developing liver fibrosis, cirrhosis, liver cancer and other serious complications in the later stage. For example, in some large physical examination centers, although liver color Doppler ultrasound is used for preliminary screening, due to the problem of missed diagnosis, some patients are diagnosed only after they have more serious symptoms of liver disease in the future, and the difficulty and cost of treatment are greatly increased at this time.

[0004] At present, in the field of fatty liver diagnosis, there is no effective method to identify patients who are missed by liver color Doppler ultrasound examination. This situation seriously restricts the early diagnosis and treatment of fatty liver. With the continuous development of medical technology, people have higher and higher requirements for the accuracy and timeliness of disease diagnosis. For a disease with a high prevalence and serious harm such as NAFLD, a reliable method is urgently needed to make up for the shortcomings of existing screening methods. Therefore, it is urgent to establish a method to identify patients with fatty liver who are missed by liver color Doppler ultrasound examination to prevent the missed diagnosis of fatty liver in health examinations. After reviewing the literature, we found that no relevant method has been established yet. Summary of the invention

[0005] In view of the above problems, a method and system for detecting patients with fatty liver missed by liver color Doppler ultrasound examination is provided. By using the logistic regression algorithm and combining other related technologies and data, a method is established to accurately identify patients with fatty liver missed by liver color Doppler ultrasound, thereby achieving early detection, early diagnosis and early treatment of people with fatty liver. This is of great significance for reducing the disease burden of NAFLD and improving the quality of life and prognosis of patients.

[0006] In order to achieve the above-mentioned object, as a first aspect of the present invention, a method for detecting patients with fatty liver missed by liver color Doppler ultrasound examination is provided, comprising the following steps: Step 1: Data collection: Collect sample data of the examinees including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire data; Step 2: Data cleaning: Perform data cleaning on the sample data obtained in step 1, including deleting missing values ​​and excluding non-target physical examination data; Step 3: Preliminary model establishment: Establish a preliminary binary logistic regression model based on all variables; Step 4: Variable screening: Combining multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, the comprehensive importance score is calculated by considering the frequency of variables selected in different simplification methods and the importance weight of medical profession, and the most valuable variables for the model are screened from all variables; Step 5: Establish a predictive diagnostic model based on the logistic regression algorithm: Based on the variables screened in step 4, perform variable coding, and use fatty liver diagnosed by FibroTouch as the outcome variable to establish a variable diagnostic model based on the logistic regression algorithm; the diagnostic model identifies those who are not diagnosed with fatty liver in liver color Doppler ultrasound examination, and outputs whether they have non-alcoholic fatty liver as the diagnosis result.

[0007] Furthermore, the sample data set in step one is divided into a training set, an internal test set and an external test set; the internal test set and the external test set are sample data collected from two different hospital districts; the internal test set and the training set are data collected from the same hospital district.

[0008] Furthermore, the method of excluding non-target physical examination data in step 2 includes: excluding patients diagnosed with fatty liver by liver color Doppler ultrasound; excluding those with CAP quartile greater than 30db / m; excluding those who self-reported alcoholic liver disease; excluding those with viral hepatitis; excluding those with schistosomiasis liver disease; and excluding pregnant women.

[0009] Furthermore, all variables in step three include age, gender, education level, place of residence, smoking, drinking, regular physical exercise, sleep duration at night, nap, daily vegetable intake, daily fruit intake, daily red meat intake, systolic blood pressure, diastolic blood pressure, hypertension, body mass index, fasting blood sugar, diabetes, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, triglycerides, blood uric acid, blood creatinine, alanine aminotransferase, aspartate aminotransferase, gamma-glutamyl transpeptidase, and fatty liver diagnosed based on FibroTouch.

[0010] Furthermore, the variables screened in step 4 include gender, age, low-density lipoprotein cholesterol, triglycerides, gamma-glutamyl transpeptidase, blood uric acid, regular physical exercise and body mass index; before screening, no less than 25 characteristic variables are initially included in the model; after screening, no more than 8 variables are finally included in the model.

[0011] Furthermore, the step 4 specifically includes the following steps: (1) Combination of multiple methods First, the logistic regression backward method, stepwise regression method and ridge regression method are used to screen the variables of the original model. The logistic regression backward method starts from the full model containing all variables, and removes one variable that contributes the least and is not significant to the model each time until the preset stop condition is reached. The stepwise regression method gradually introduces and removes variables to determine the final variable set according to the contribution of the variables to the model. The ridge regression method adds a regularization term to the estimation of the regression coefficient to prevent overfitting and screen the variables at the same time. Three sets of simplified variable sets are obtained through these three methods. (2) Comprehensive evaluation of variable importance A comprehensive evaluation was conducted on the variable sets obtained by the above three methods, and the frequency of each variable being selected in different methods was calculated, and the frequency was used as a measure of the importance of the variable. At the same time, the professional importance of the variable in the medical field was considered. With reference to medical literature and professional knowledge, the relevance of each variable to fatty liver disease was evaluated and different weights were given. For example, some key variables that have been confirmed to be closely related to fatty liver by a large number of medical studies, such as triglycerides and body mass index, were given higher weights. According to the frequency of variable selection and the importance weight of medical profession, the comprehensive importance score of each variable is calculated. The calculation formula of the comprehensive importance score is:

[0012] in, Indicates The comprehensive importance score of the variables, Indicates The frequency with which a variable is selected in the three simplification methods, Indicates The weights of the variables are determined based on medical expertise; (3) Determine the final variable set According to the comprehensive importance score of the variables, the variables with scores higher than the preset threshold are selected as the variables finally included in the model; the preset threshold can be adjusted according to the actual situation and the performance requirements of the industry model; in this way, the final set of variables is determined and the simplified model is established.

[0013] Furthermore, the variable coding method in step 5 is as follows: age, years old; gender, male=1, female=0; education level, high school and below=0, university / college=1, graduate and above=2; place of residence, urban=1, rural=0; smoking, yes=1, no=0; drinking, yes=1, no=0; regular physical exercise, yes=1, no=0; sleep duration at night, hours; nap, yes=1, no=0; daily vegetable intake, yes=1, no=0; daily fruit intake, yes=1, no=0; daily red meat intake, yes=1, no=0; systolic blood pressure, mmHg; diastolic blood pressure, mmHg; hypertension, yes=1, no=0; body mass index, kg / m 2 ; Fasting blood glucose, mmol / L; Diabetes, yes=1, no=0; Low-density lipoprotein cholesterol, mmol / L; High-density lipoprotein cholesterol, mmol / L; Total cholesterol, mmol / L; Triglycerides, mmol / L; Serum uric acid, umol / L; Serum creatinine, umol / L; Alanine aminotransferase, U / L; Aspartate aminotransferase, U / L; γ-glutamyl transpeptidase, U / L; Fatty liver diagnosed based on FibroTouch, yes=1, no=0.

[0014] As a second aspect of the present invention, a system for detecting patients with fatty liver who are missed by liver color Doppler ultrasound examination through logistic regression is also provided, comprising: a data collection unit, a data cleaning unit, a preliminary model establishment unit, a variable screening unit, and a predictive diagnosis model establishment unit based on a logistic regression algorithm; The data collection unit is used to collect sample data of the examinee including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire survey data; The data cleaning unit is used to clean the sample data, including deleting missing values ​​and excluding non-target physical examination data; The preliminary model building unit is used to build a preliminary binary logistic regression model based on all variables; The variable screening unit is used to combine multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, consider the frequency of variables selected in different simplification methods and the importance weight of medical profession to calculate the comprehensive importance score, and screen the most valuable variables for the model from all variables; The predictive diagnosis model unit based on the logistic regression algorithm is used to encode variables based on the screened variables, and to establish a variable diagnosis model based on the logistic regression algorithm with fatty liver diagnosed based on FibroTouch as the outcome variable; the diagnostic model identifies those who are not diagnosed with fatty liver in liver color ultrasound examination, and outputs whether they have non-alcoholic fatty liver as the diagnosis result.

[0015] As a third aspect of the present invention, a computer-readable storage medium is further provided, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method steps for detecting patients with fatty liver missed by liver color ultrasound examination as described above are implemented.

[0016] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art: (1) The method of detecting patients with fatty liver who are missed by liver color Doppler ultrasound examination of the present invention establishes a multi-level fatty liver screening system. The first screening procedure is liver color Doppler ultrasound examination. Those who are not diagnosed with fatty liver enter the second screening procedure, that is, the computer system automatically extracts the variables such as gender, age, low-density lipoprotein cholesterol, triglycerides, γ-glutamyl transpeptidase, blood uric acid, regular physical exercise and body mass index of the examinee, and substitutes them into our diagnostic model, and outputs the diagnosis result whether it is non-alcoholic fatty liver. Effectively identify patients with non-alcoholic fatty liver who are missed by liver color Doppler ultrasound, and through early intervention, avoid their progression to liver fibrosis, cirrhosis, liver cancer, cardiovascular and cerebrovascular diseases, chronic kidney disease and other consequences.

[0017] (2) The method of the present invention for detecting patients with fatty liver who have been missed by color Doppler ultrasonography of the liver, by combining a variety of methods (e.g., logistic regression, stepwise regression, and ridge regression), takes into account the frequency of variables being selected in different simplification methods and the importance weights of medical professions to calculate a comprehensive importance score, thereby achieving a more scientific and reasonable technical effect of determining the degree of contribution of the variables to the model, avoiding the problem that a single method may lose useful information, and improving the model's predictive ability for different situations and subgroups.

[0018] (3) The method of the present invention for detecting patients with fatty liver who have been missed by color Doppler ultrasound of the liver integrates multi-dimensional data such as color Doppler ultrasound of the liver, FibroTouch testing, biochemical testing, physical measurements, and questionnaires. This comprehensive data collection method breaks the limitations of traditional single testing methods and can obtain the health information of the examinee from multiple perspectives, providing a rich data basis for accurate diagnosis of liver health status. For example, through various indicators in biochemical testing such as triglycerides and cholesterol, combined with data such as body mass index from physical measurements, as well as questionnaire information related to lifestyle habits (such as smoking, drinking, physical exercise, etc.), a more comprehensive understanding of the physical condition of the examinee and the impact of lifestyle on liver health can be obtained, thereby more accurately judging the potential risk of fatty liver. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a method for detecting patients with fatty liver who were missed by liver color Doppler ultrasonography based on a logistic regression algorithm in a preferred embodiment; Figure 2 A receiver operating characteristic (ROC) curve diagram of a full variable logistic regression diagnostic model based on a training set in a preferred embodiment; Figure 3 A receiver operating characteristic (ROC) curve diagram of the final diagnostic model based on the training set of a preferred embodiment; Figure 4 A receiver operating characteristic (ROC) curve diagram of the final diagnostic model based on the internal test set of a preferred embodiment; Figure 5 A receiver operating characteristic (ROC) curve diagram of the final diagnostic model based on the external test set of a preferred embodiment; Figure 6 Schematic diagram of system units of a preferred embodiment. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 Please refer to Figure 1 This embodiment 1 provides a method for diagnosing patients with fatty liver who are missed by liver color Doppler ultrasound examination during a health check-up, comprising: Step 1: Data collection: Collect sample data of the examinees including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire data; The ultrasonic attenuation parameter (CAP) of FibroTouch, a transient elastography technology, is more sensitive than ultrasound imaging in diagnosing fatty liver, and can detect fatty liver with >5% fatty degeneration. CAP as a continuous variable can also be used to monitor changes in liver fat content. According to the diagnostic criteria of the "Guidelines for the Prevention and Treatment of Metabolic (Non-alcoholic) Fatty Liver Disease (2024 Edition)", fatty liver can be diagnosed when CAP ≥ 244. FibroTouch is also an instrument for assessing the degree of liver fibrosis. It is not as widely used in health examinations as liver color ultrasound, and only a small number of examinees complete this examination.

[0022] The subjects included in this method are 16,632 people who completed physical examinations such as liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement, and questionnaire survey in a certain health management center (Campus 1 and Campus 2) from November 2021 to February 2024.

[0023] In other implementations, a combination of multimodal detection methods has been carried out. In addition to FibroTouch technology and liver color Doppler ultrasound, other liver detection technologies can be combined, such as magnetic resonance imaging (MRI), positron emission tomography (PET), etc. Different detection technologies have different advantages and characteristics. MRI has high accuracy and resolution in detecting liver fat content, and PET can reflect the metabolic function of the liver. By integrating the data of multiple detection technologies and establishing a multimodal diagnostic model, fatty liver can be evaluated from multiple angles to improve the accuracy and reliability of diagnosis. At the same time, for some difficult cases or suspected patients with fatty liver, more comprehensive diagnostic information can be provided to avoid misdiagnosis and missed diagnosis.

[0024] Step 2: Data cleaning: Perform data cleaning on the sample data obtained in step 1, including deleting missing values ​​and excluding non-target physical examination data; In this embodiment, among the 16,632 people, 6,913 patients diagnosed with fatty liver by liver ultrasound were first excluded, and then 1,084 people including those with CAP quartile>30db / m, those who reported alcoholic liver disease in the questionnaire, those with viral hepatitis, those with schistosomiasis liver disease, and pregnant women were successively eliminated, leaving 8,635 people. This method initially included 28 characteristic variables, and after deleting missing values, 8,156 people were finally included in the analysis, including 5,429 people in the Hankou campus and 2,727 people in the Optics Valley campus.

[0025] Step 3: Preliminary model establishment: Establish a preliminary binary logistic regression model based on all variables: A full-variable diagnostic model based on the logistic regression algorithm was established: the variables and codes included in this method were: age (years), sex (male = 1, female = 0), education level (high school and below = 0, university / college = 1, graduate and above = 2), place of residence (urban = 1, rural = 0), smoking (yes = 1, no = 0), drinking (yes = 1, no = 0), regular physical exercise (yes = 1, no = 0), sleep duration at night (hours), nap (yes = 1, no = 0), daily vegetable intake (yes = 1, no = 0), daily fruit intake (yes = 1, no = 0), daily red meat intake (yes = 1, no = 0), systolic blood pressure (mmHg) , diastolic blood pressure (mmHg), hypertension (yes = 1, no = 0), body mass index (kg / m2), fasting blood glucose (mmol / L), diabetes (yes = 1, no = 0), low-density lipoprotein cholesterol (mmol / L), high-density lipoprotein cholesterol (mmol / L), total cholesterol (mmol / L), triglycerides (mmol / L), blood uric acid (umol / L), blood creatinine (umol / L), alanine aminotransferase (U / L), aspartate aminotransferase (U / L), γ-glutamyl transpeptidase (U / L), fatty liver diagnosed by FibroTouch (yes = 1, no = 0). With fatty liver as the binary outcome variable and the other 27 characteristic variables as predictor variables, 5429 physical examination subjects in Hankou campus were randomly divided into training set (4343 people) and internal test set (1086 people) in a ratio of 8:2, and 2714 physical examination subjects in Optical Valley campus were used as external test set. A binary logistic regression model was established based on the training set data. The specific parameters are shown in formula (1): Ln[P / (1-P)]=0.210×sex+0.014×age+0.089×low-density lipoprotein cholesterol+0.225×triglycerides-0.158×high-density lipoprotein cholesterol-0.003×total cholesterol+0.073×fasting blood glucose+0.005×alanine aminotransferase-0.009×aspartate aminotransferase+0.003×γ-glutamyl transpeptidase-0.002×serum creatinine+0.002×serum uric acid+0.099×residence-0.106×smoking+0.094 × drinking + 0.041 × daily fruit intake + 0.051 × daily vegetable intake + 0.151 × daily red meat intake - 0.128 × regular physical exercise - 0.060 × nap - 0.065 × nighttime sleep duration - 0.306 × diabetes + 0.004 × systolic blood pressure - 0.003 × diastolic blood pressure + 0.458 × body mass index - 0.018 × hypertension + 0.019 × education level (university / college vs high school and below) + 0.023 × education level (graduate and above vs high school and below) - 12.590 (1) In formula (1), P represents the probability of diagnosing missed non-alcoholic fatty liver disease. The model showed excellent diagnostic performance, with an area under the curve (AUC) (95% CI) of 0.829 (0.818-0.840) and a receiver operating characteristic (ROC) curve of Figure 2 .

[0026] For the data in this case, continuous variables can be expanded polynomially. For example, for the feature of age, in addition to using the original age value, you can also add the square term, cubic term, etc. of age to capture the possible nonlinear relationship between age and fatty liver. In addition, principal component analysis (PCA) can be performed to convert multiple related original features into a few unrelated principal components. For example, for multiple blood biochemical indicators (such as triglycerides, low-density lipoprotein cholesterol, total cholesterol, etc.), PCA can combine them into new principal components that can retain the variance information of the original data to the greatest extent.

[0027] Step 4: Variable screening: Combining multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, the comprehensive importance score is calculated by considering the frequency of variables being selected in different simplification methods and the importance weight of medical profession, and the most valuable variables for the model are screened from all variables.

[0028] In this embodiment, model variables are screened based on the logistic regression backward method: Since the above-mentioned full variable model has 27 predictive variables, it is difficult to implement in practical applications. Therefore, the logistic regression backward method is used to screen the model variables, so as to achieve the purpose of streamlining the model to avoid collinearity and overfitting. Figure 2 This is a step diagram for selecting variables using the logistic regression method. After screening, the variables finally included in the model are gender, age, low-density lipoprotein cholesterol, triglycerides, γ-glutamyl transpeptidase, blood uric acid, regular physical exercise, and body mass index. The model parameters are shown in formula (2): Ln[P / (1-P)]=0.202×sex+0.013×age+0.095×low-density lipoprotein cholesterol0.241×triglyceride+0.003×γ-glutamyl transpeptidase+0.002×serum uric acid-0.133×regular physical exercise+0.467×body mass index-12.893 (2) In other preferred embodiments, step 4 further includes: (1) Combination of multiple methods First, the logistic regression backward method, stepwise regression method and ridge regression method are used to screen the variables of the original model. The logistic regression backward method starts from the full model containing all variables, and removes one variable that contributes the least and is not significant to the model each time until the preset stop condition is reached. The stepwise regression method gradually introduces and removes variables to determine the final variable set according to the contribution of the variables to the model. The ridge regression method adds a regularization term to the estimation of the regression coefficient to prevent overfitting and screen the variables at the same time. Three sets of simplified variable sets are obtained through these three methods. (2) Comprehensive evaluation of variable importance A comprehensive evaluation was conducted on the variable sets obtained by the above three methods, and the frequency of each variable being selected in different methods was calculated, and the frequency was used as a measure of the importance of the variable. At the same time, the professional importance of the variable in the medical field was considered. With reference to medical literature and professional knowledge, the relevance of each variable to fatty liver disease was evaluated and different weights were given. For example, some key variables that have been confirmed to be closely related to fatty liver by a large number of medical studies, such as triglycerides and body mass index, were given higher weights. According to the frequency of variable selection and the importance weight of medical profession, the comprehensive importance score of each variable is calculated. The calculation formula of the comprehensive importance score is:

[0029] in, Indicates The comprehensive importance score of the variables, Indicates The frequency with which a variable is selected in the three simplification methods, Indicates The weights of the variables are determined based on medical expertise; (3) Determine the final variable set According to the comprehensive importance score of the variables, the variables with scores higher than the preset threshold are selected as the variables finally included in the model; the preset threshold can be adjusted according to the actual situation and the performance requirements of the industry model; in this way, the final set of variables is determined and the simplified model is established.

[0030] Step 5: Establish a predictive diagnostic model based on the logistic regression algorithm: Based on the variables screened in step 4, perform variable coding, and use fatty liver diagnosed by FibroTouch as the outcome variable to establish a variable diagnostic model based on the logistic regression algorithm; the diagnostic model identifies those who are not diagnosed with fatty liver in liver color Doppler ultrasound examination, and outputs whether they have non-alcoholic fatty liver as the diagnosis result.

[0031] Establish the final logistic regression algorithm diagnostic model: Train and establish the logistic regression diagnostic model according to the 10-fold cross-validation mode. Based on the logistic regression algorithm, gender, age, low-density lipoprotein cholesterol, triglycerides, γ-glutamyl transpeptidase, serum uric acid, regular physical exercise and body mass index were used as independent variables, and fatty liver diagnosed by FibroTouch was used as the outcome variable. The final diagnostic model was established and verified on the internal test set and the external test set. The model diagnostic performance was excellent. The specific indicator parameters are shown in Table 1. The ROC curves on the training set, internal test set, and external test set are shown in Table 1. Figure 2 , Figure 3 , Figure 4 The areas under the curve are 0.823, 0.821, and 0.810 respectively.

[0032] Table 1 Characteristics of the diagnostic model for non-alcoholic fatty liver disease missed by liver ultrasound based on the logistic regression algorithm

[0033] In other embodiments, an auxiliary software or system specifically used for the diagnosis of fatty liver is also involved. The software or system can integrate functions such as data acquisition, model calculation, and output of diagnostic results. In a physical examination center or medical institution, the doctor only needs to input the patient's relevant data into the system, and the system can automatically calculate according to the logistic regression algorithm model of this patent, and quickly output the diagnostic results and risk assessment report. At the same time, the system can also provide a visual interface, such as drawing a receiver operating characteristic (ROC) curve, etc., to facilitate doctors to intuitively understand the performance of the diagnostic model and the patient's condition.

[0034] The software or system can also have data management and analysis functions, and can store and analyze a large amount of physical examination data, and provide support for medical institutions in epidemiological research, disease trend analysis, etc. For example, by analyzing the data of fatty liver patients in different time periods and regions, we can understand the trend of fatty liver disease and the changes in risk factors, and provide a basis for the formulation of public health policies and the allocation of medical resources.

[0035] Example 2 Please refer to Figure 6 This embodiment 2 provides a system for diagnosing patients with fatty liver who are missed by liver color Doppler ultrasound examination during a health check-up, including: A data collection unit is used to collect sample data of the examinees, including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire data; A data cleaning unit is used to clean the sample data, including deleting missing values ​​and excluding non-target physical examination data; A preliminary model building unit is used to build a preliminary binary logistic regression model based on all variables; The variable screening unit is used to combine multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, consider the frequency of variables selected in different simplification methods and the importance weight of medical profession to calculate the comprehensive importance score, and screen the most valuable variables for the model from all variables; A predictive diagnostic model unit based on the logistic regression algorithm was established to encode variables based on the screened variables, and a variable diagnostic model based on the logistic regression algorithm was established with fatty liver diagnosed based on FibroTouch as the outcome variable; the diagnostic model identified those who were not diagnosed with fatty liver in liver color Doppler ultrasound examination, and output whether they had non-alcoholic fatty liver as the diagnosis result.

[0036] Example 3 This embodiment 3 also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a method for diagnosing patients with fatty liver who were missed by color Doppler ultrasound of the liver during a health checkup can be implemented.

[0037] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0038] For an introduction to the computer-readable storage medium provided in this application, please refer to the above method embodiment, and this application will not go into details here.

[0039] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting patients with fatty liver who are missed by liver color Doppler ultrasound examination, characterized in that: The following steps are involved: Step 1: Data collection: Collect sample data of the examinees including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire data; Step 2: Data cleaning: Perform data cleaning on the sample data obtained in step 1, including deleting missing values ​​and excluding non-target physical examination data; Step 3: Preliminary model establishment: Establish a preliminary binary logistic regression model based on all variables; Step 4: Variable screening: Combining multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, the comprehensive importance score is calculated by considering the frequency of variables selected in different simplification methods and the importance weight of medical profession, and the most valuable variables for the model are screened from all variables; Step 5: Establish a predictive diagnostic model based on the logistic regression algorithm: Based on the variables screened in step 4, perform variable coding, and use fatty liver diagnosed by FibroTouch as the outcome variable to establish a variable diagnostic model based on the logistic regression algorithm; the diagnostic model identifies those who are not diagnosed with fatty liver in liver color Doppler ultrasound examination, and outputs whether they have non-alcoholic fatty liver as the diagnosis result; The steps 1 to 5 are all processed and completed by a computer device.

2. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: The sample data set in step 1 is divided into a training set, an internal test set and an external test set; the internal test set and the external test set are sample data collected from two different hospital districts; the internal test set and the training set are data collected from the same hospital district; The data set is divided and processed by computer means.

3. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: The method for excluding non-target physical examination data in step 2 includes: Information processing is performed by computer devices to complete the following steps: Patients diagnosed with fatty liver by liver ultrasound were excluded; Those with CAP quartile greater than 30db / m were excluded; Those who self-reported alcoholic liver disease were excluded; Patients with viral hepatitis were excluded; Patients with schistosomiasis liver disease were excluded; Pregnant women were excluded.

4. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: All variables in step three include age, gender, education level, place of residence, smoking, drinking, regular physical exercise, sleep duration at night, nap, daily vegetable intake, daily fruit intake, daily red meat intake, systolic blood pressure, diastolic blood pressure, hypertension, body mass index, fasting blood sugar, diabetes, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, triglycerides, blood uric acid, blood creatinine, alanine aminotransferase, aspartate aminotransferase, gamma-glutamyl transpeptidase, and fatty liver diagnosed based on FibroTouch.

5. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: The variables screened in step 4 include gender, age, low-density lipoprotein cholesterol, triglycerides, gamma-glutamyl transpeptidase, blood uric acid, regular physical exercise and body mass index; before screening, no less than 25 characteristic variables are initially included in the model; after screening, no more than 8 variables are finally included in the model, and the screening process is completed by processing through a computer device.

6. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: The step 4 specifically includes the following steps: Use a computer device to complete the following specific steps: (1) Combination of multiple methods First, the logistic regression backward method, stepwise regression method and ridge regression method are used to screen the variables of the original model. The logistic regression backward method starts from the full model containing all variables, and removes one variable that contributes the least and is not significant to the model each time until the preset stop condition is reached. The stepwise regression method gradually introduces and removes variables to determine the final variable set according to the contribution of the variables to the model. The ridge regression method adds a regularization term to the estimation of the regression coefficient to prevent overfitting and screen the variables at the same time. Three sets of simplified variable sets are obtained through these three methods. (2) Comprehensive evaluation of variable importance A comprehensive evaluation was conducted on the variable sets obtained by the above three methods, and the frequency of each variable being selected in different methods was calculated, and the frequency was used as a measure of the importance of the variable. At the same time, the professional importance of the variable in the medical field was considered. With reference to medical literature and professional knowledge, the relevance of each variable to fatty liver disease was evaluated and different weights were given. For example, some key variables that have been confirmed to be closely related to fatty liver by a large number of medical studies, such as triglycerides and body mass index, were given higher weights. According to the frequency of variable selection and the importance weight of medical profession, the comprehensive importance score of each variable is calculated. The calculation formula of the comprehensive importance score is: ， in, Indicates The comprehensive importance score of the variables, Indicates The frequency with which a variable is selected in the three simplification methods, Indicates The weights of the variables are determined based on medical expertise; (3) Determine the final variable set According to the comprehensive importance score of the variables, the variables with scores higher than the preset threshold are selected as the variables finally included in the model; the preset threshold can be adjusted according to the actual situation and the performance requirements of the industry model; in this way, the final set of variables is determined and the simplified model is established.

7. The method for detecting patients with fatty liver missed by liver color Doppler ultrasonography according to claim 1, characterized in that: The variable coding method in step 5 is: age, years old; Gender, male = 1, female = 0; education level, high school or below = 0, college / junior college = 1, graduate or above = 2; place of residence, urban = 1, rural = 0; smoking, yes = 1, no = 0; Drinking, yes = 1, no = 0; Regular physical exercise, yes = 1, no = 0; Nighttime sleep duration, hours; Nap, yes=1, no=0; daily vegetable intake, yes=1, no=0; daily fruit intake, yes=1, no=0; daily red meat intake, yes=1, no=0; systolic blood pressure, mmHg; diastolic blood pressure, mmHg; hypertension, yes=1, no=0; body mass index, kg / m 2 ; Fasting blood glucose, mmol / L; Diabetes, yes=1, no=0; Low-density lipoprotein cholesterol, mmol / L; High-density lipoprotein cholesterol, mmol / L; Total cholesterol, mmol / L; Triglycerides, mmol / L; Serum uric acid, umol / L; Serum creatinine, umol / L; Alanine aminotransferase, U / L; Aspartate aminotransferase, U / L; γ-glutamyl transpeptidase, U / L; Fatty liver diagnosed based on FibroTouch, yes=1, no=0.

8. A system for detecting patients with fatty liver who are missed by liver color Doppler ultrasound examination using logistic regression, characterized in that: include: Data collection unit, data cleaning unit, preliminary model building unit, variable screening unit, and predictive diagnosis model building unit based on logistic regression algorithm; The data collection unit is used to collect sample data of the examinee including liver color Doppler ultrasound examination, FibroTouch test, biochemical test, physical measurement and questionnaire survey data; The data cleaning unit is used to clean the sample data, including deleting missing values ​​and excluding non-target physical examination data; The preliminary model building unit is used to build a preliminary binary logistic regression model based on all variables; The variable screening unit is used to combine multiple methods including logistic regression backward method, stepwise regression method and ridge regression method, consider the frequency of variables selected in different simplification methods and the importance weight of medical profession to calculate the comprehensive importance score, and screen the most valuable variables for the model from all variables; The predictive diagnosis model unit based on the logistic regression algorithm is used to encode variables based on the screened variables, and to establish a variable diagnosis model based on the logistic regression algorithm with fatty liver diagnosed based on FibroTouch as the outcome variable; the diagnostic model identifies those who are not diagnosed with fatty liver in liver color ultrasound examination, and outputs whether they have non-alcoholic fatty liver as the diagnosis result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps for detecting patients with fatty liver that are missed by liver color Doppler ultrasound examination as described in any one of claims 1 to 7 are implemented.

Images