Method, system, equipment and medium for constructing diagnostic model of non-alcoholic fatty liver disease

Abstract

The present application provides a method, system, device and medium for constructing a non-alcoholic fatty liver disease diagnostic model. The method includes obtaining first user information from a first data set; screening the first user information, screening the gene information according to minimum absolute value shrinkage and selection operator regression to obtain risk gene information; constructing a non-alcoholic fatty liver disease diagnostic model, the non-alcoholic fatty liver disease diagnostic model includes a basic module and a meta-module, obtaining a receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnostic model, if the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnostic model can be put into use, combining a variety of advanced machine learning algorithms to enhance the generalization ability and data processing ability of the model, and having diagnostic accuracy and reliability in NAFLD diagnosis compared to the prior art.

Description

Technical Field

[0001] The present application relates to the field of health diagnosis technology, and in particular to a method, system, device and medium for constructing a diagnosis model for non-alcoholic fatty liver disease. Background Art

[0002] Non-alcoholic fatty liver disease (NAFLD) has become the most common chronic liver disease in the past four decades, with a global adult prevalence of approximately 25%. In China, its prevalence is as high as nearly 30%. This high prevalence has caused NAFLD to become the main cause of the rapid increase in liver-related mortality worldwide. In addition, NAFLD is not just a liver problem, it is closely associated with the risk of a variety of serious complications, including cardiovascular disease, type 2 diabetes, chronic kidney disease, etc. As the disease progresses, NAFLD may also develop into end-stage liver disease, primary liver cancer, and the need for liver transplantation, placing a heavy burden on the health system and the economy. Therefore, the effective prevention, diagnosis and management of NAFLD is not only crucial to individual health, but also of great significance to global public health.

[0003] Currently, a variety of biomarker models have been applied to the diagnosis of NAFLD, such as enhanced liver fibrosis (ELF), fibrosis-4 index (FIB-4) and non-alcoholic fatty liver fibrosis score (NAFLD fibrosis score, NFS). However, these methods are greatly affected by the operator's level and ability, have high costs, limited applicability, and accuracy that needs to be improved. They also mainly rely on clinical indicators. Summary of the invention

[0004] In view of this, it is necessary to provide a method, system, device and medium for constructing a diagnostic model of non-alcoholic fatty liver disease that can overcome at least one of the above defects.

[0005] In a first aspect, an embodiment of the present application provides a method for constructing a non-alcoholic fatty liver disease diagnosis model, comprising: obtaining first user information from a first data set; screening the first user information to obtain clinical information and gene information in the first user information; obtaining clinical information and gene information in the training set and the internal test set; obtaining risk clinical information in the clinical information; screening the gene information according to minimum absolute value shrinkage and selection operator regression to obtain risk gene information; constructing a non-alcoholic fatty liver disease diagnosis model, the non-alcoholic fatty liver disease diagnosis model comprising a basic module and a meta-module, wherein the basic module and the meta-module are connected in a cascade manner, the basic module comprises a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression and K nearest neighbor, and the meta-module comprises a logistic regression classifier; training the non-alcoholic fatty liver disease diagnosis model according to the risk clinical information and the risk gene information; obtaining a receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model, if the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use.

[0006] According to one embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnostic model further includes: evaluating the performance of the non-alcoholic fatty liver disease diagnostic model according to a five-fold cross-validation method.

[0007] According to one embodiment of the present application, the performance of the non-alcoholic fatty liver disease diagnosis model is evaluated according to the five-fold cross-validation method, including: randomly dividing the screened risk clinical information and the risk gene information into five equal parts, four of which are used as training sets and one is used as an internal validation set; in each iteration, the non-alcoholic fatty liver disease diagnosis model is trained according to the four training sets, and the non-alcoholic fatty liver disease diagnosis model is validated according to the internal validation set, and the performance index between the predicted result of the non-alcoholic fatty liver disease diagnosis model and the actual label is calculated each time until each of the training sets is used as the internal validation set; the performance index in all iterations is obtained, and the model structure and parameters of the non-alcoholic fatty liver disease diagnosis model are adjusted according to the performance index in all iterations until the optimal non-alcoholic fatty liver disease diagnosis model is obtained.

[0008] According to one embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnostic model also includes: the non-alcoholic fatty liver disease diagnostic model is used to output outcome events, and the outcome events include adverse intrahepatic outcomes and adverse extrahepatic outcomes; the adverse intrahepatic outcomes include that individuals with non-alcoholic fatty liver disease also develop severe liver disease; the adverse extrahepatic outcomes include that individuals with non-alcoholic fatty liver disease also develop hypertension, type 2 diabetes, chronic kidney disease, coronary heart disease, heart failure, and atrial fibrillation or atrial flutter.

[0009] According to one embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnosis model further includes: obtaining second user information from a second data set, and screening the second user information to obtain an external test set. According to one embodiment of the present application, the internal test set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnosis model on processing training data, and the external test set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnosis model on processing unseen data.

[0010] According to one embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnostic model further includes: using grid search or Bayesian optimization to determine the optimal hyperparameters of the non-alcoholic fatty liver disease diagnostic model.

[0011] In a second aspect, an embodiment of the present application provides a system for constructing a diagnostic model for non-alcoholic fatty liver disease, which is applied to the method for constructing a diagnostic model for non-alcoholic fatty liver disease as described in the first aspect, including: an information acquisition module, used to obtain first user information from a first data set; an information screening module, used to screen the first user information to obtain clinical information and gene information in the first user information; an information acquisition module, used to obtain clinical information and gene information in the training set and the internal test set, obtain risk clinical information in the clinical information, and screen the gene information according to minimum absolute value shrinkage and selection operator regression to obtain risk gene information; a model construction module, used to construct a non-alcoholic fatty liver disease diagnostic model. An alcoholic fatty liver disease diagnosis model, the non-alcoholic fatty liver disease diagnosis model includes a basic module and a meta-module, wherein the basic module and the meta-module are connected in a cascade manner, the basic module includes a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression and K nearest neighbor, and the meta-module includes a logistic regression classifier; a model training module is used to train the non-alcoholic fatty liver disease diagnosis model according to the risk clinical information and the risk gene information, and obtain the receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model. If the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use.

[0012] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to implement the method for constructing a non-alcoholic fatty liver disease diagnostic model described in the first aspect when executing the instructions.

[0013] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, wherein the instructions instruct a device to execute the method for constructing a non-alcoholic fatty liver disease diagnostic model as described in the first aspect.

[0014] The method, system, device and medium for constructing a non-alcoholic fatty liver disease diagnostic model provided in the embodiments of the present application first screen the first user information to extract key clinical and genetic information related to NAFLD, then apply minimum absolute value shrinkage and selection operator regression to perform gene screening to accurately capture risk gene information, and combine a variety of advanced machine learning algorithms to enhance the generalization and data processing capabilities of the model, which has diagnostic accuracy and reliability in NAFLD diagnosis compared to the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic flow chart of a method for constructing a non-alcoholic fatty liver disease diagnostic model provided in one embodiment of the present application.

[0016] Figure 2 A schematic diagram of a system for constructing a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application.

[0017] Figure 3 A block diagram of a non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0018] Figure 4a This is a predictive performance graph of the non-alcoholic fatty liver disease diagnostic model provided in one embodiment of the present application.

[0019] Figure 4b This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application.

[0020] Figure 4c This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application.

[0021] Figure 4d This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application.

[0022] Figure 4e This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application.

[0023] Figure 5aThis is a prediction performance graph of the non-alcoholic fatty liver disease diagnosis model training set provided in one embodiment of the present application.

[0024] Figure 5b This is a prediction performance graph of a non-alcoholic fatty liver disease diagnosis model training set provided by another embodiment of the present application.

[0025] Figure 6 This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0026] Figure 7a This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0027] Figure 7b This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0028] Figure 8 A schematic diagram of an electronic device provided in one embodiment of the present application.

[0029] Main component symbols

[0030] Non-alcoholic fatty liver disease diagnostic model construction system 10

[0031] Information collection module 110

[0032] Information filtering module 120

[0033] Information acquisition module 130

[0034] Model building module 140

[0035] Model training module 150

[0036] Electronic devices 20

[0037] Processor 21

[0038] Memory 22

[0039] Nonalcoholic fatty liver disease diagnostic model 30

[0040] Basic Module 31

[0041] Meta module 32

[0042] Method steps S100-S800 DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0044] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0045] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0046] Based on the implementations in this application, all other implementations obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] Non-alcoholic fatty liver disease (NAFLD) has become the most common chronic liver disease in the past four decades, with a global adult prevalence of approximately 25%. In China, its prevalence is as high as nearly 30%. This high prevalence has caused NAFLD to become the main cause of the rapid increase in liver-related mortality worldwide. In addition, NAFLD is not just a liver problem, it is closely associated with the risk of a variety of serious complications, including cardiovascular disease, type 2 diabetes, chronic kidney disease, etc. As the disease progresses, NAFLD may also develop into end-stage liver disease, primary liver cancer, and the need for liver transplantation, placing a heavy burden on the health system and the economy. Therefore, the effective prevention, diagnosis and management of NAFLD is not only crucial to individual health, but also of great significance to global public health.

[0048] Currently, a variety of biomarker models have been applied to the diagnosis of NAFLD, such as enhanced liver fibrosis (ELF), fibrosis-4 index (FIB-4) and non-alcoholic fatty liver fibrosis score (NAFLD fibrosis score, NFS). However, these methods are greatly affected by the operator's level and ability, have high costs, limited applicability, and accuracy that needs to be improved. They also mainly rely on clinical indicators.

[0049] Therefore, the embodiments of the present application provide a method, system, device and medium for constructing a non-alcoholic fatty liver disease diagnostic model. First, by screening the first user information, key clinical and genetic information related to NAFLD is extracted, and then minimum absolute value shrinkage and selection operator regression are applied for gene screening to accurately capture risk gene information. Combined with a variety of advanced machine learning algorithms, the generalization ability and data processing capabilities of the model are enhanced, which has diagnostic accuracy and reliability in NAFLD diagnosis compared with the existing technology.

[0050] In conjunction with the accompanying drawings, some embodiments of the application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0051] Figure 1 FIG. 1 is a flow chart of a method for constructing a non-alcoholic fatty liver disease diagnosis model provided in an embodiment of the present application. Figure 1 As shown, the method for constructing a diagnostic model for non-alcoholic fatty liver disease includes at least the following steps.

[0052] S100: Acquire first user information from a first data set.

[0053] In an embodiment of the present application, a method for constructing a non-alcoholic fatty liver disease diagnostic model includes obtaining first user information from a first data set.

[0054] The first dataset is, understandably, clinical and genotypic data from individuals in the UK Biobank, an ongoing prospective study that collects clinical and genotypic data from approximately 500,000 (502,389) UK individuals aged 40-69 years recruited between 2006 and 2010 and followed up multiple times.

[0055] In other embodiments, the method for constructing a non-alcoholic fatty liver disease diagnostic model further includes obtaining second user information from a second data set, and screening the second user information to obtain an external test set.

[0056] It can be understood that the second data set is data of individuals who received health examinations from a health examination center of a hospital. The data in the second data set is used as an external test set.

[0057] It is understandable that in a hospital health checkup population dataset, clinical data is derived from the physical examination electronic health record system. Given that most participants undergo continuous health checks at a hospital health management center, there are multiple blood samplings for biochemical tests. In order to improve the accuracy of the test results, each participant is matched with the biochemical analysis results closest to the time when the genetic test samples were submitted.

[0058] Specifically, individual data of participants aged ≥18 years who participated in the questionnaire survey and genetic testing were included (9,942 people in total). Subsequently, the following types of participants were excluded: participants who could not be diagnosed with fatty liver by liver ultrasound (118 people in total); participants with a history of heavy drinking or diagnosed with alcoholic fatty liver disease (30 people in total); participants with missing baseline covariate data (596 people in total); participants with missing available genotype data (367 people in total). Finally, 9,007 participants were included as an external test set.

[0059]

[0060]

[0061] Table 1: Clinical information of participants

[0062] As shown in Table 1, the number of participants in the UK Biobank was 36,490, while the number of participants in a certain hospital was 9,007. By comparing the P values, it can be seen that there were significant differences between the two groups of participants in terms of age, body mass index, triglycerides, cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, blood sugar, alanine aminotransferase, and aspartate aminotransferase (P<0.001). In addition, there was a significant difference in the incidence of non-alcoholic fatty liver disease between the two groups, with the incidence in the UK Biobank being 32.35% and the incidence in a certain hospital being 24.87% (P<0.001). These results suggest differences in clinical characteristics and the incidence of non-alcoholic fatty liver disease between the two groups of participants, providing important reference and basic data for the construction of a diagnostic model for non-alcoholic fatty liver disease.

[0063] S200: Screening the first user information to obtain clinical information and gene information in the first user information.

[0064] In an embodiment of the present application, a method for constructing a non-alcoholic fatty liver disease diagnostic model includes screening first user information to obtain clinical information and gene information in the first user information.

[0065] It is understandable that in step S200, the collected first user information is screened to extract clinical information and gene information related to the diagnosis of NAFLD. The screening process includes operations such as data cleaning, removing missing values, selecting specific clinical indicators and gene loci, etc., to ensure that the extracted information has high quality and relevance. After screening, the obtained data set will contain the key information required for building a NAFLD diagnostic model.

[0066] Specifically, when screening the first user information, participants who withdrew their consent (8 people in total), participants who could not be diagnosed with fatty liver by magnetic resonance imaging-protondensity fat fraction (MRI-PDFF) or according to the International Classification of Diseases (ICD) coding (455,218 people in total); participants with a history of heavy drinking or diagnosed with alcoholic fatty liver disease (AFLD) according to ICD (3510 people in total); ② participants with missing baseline covariate data (6796 people in total); participants with missing genotype data (367 people in total) were deleted. Finally, 36,490 user information was obtained.

[0067] In the present application embodiment, due to the lack of liver biopsy pathological data in the UK Biobank, the proton density fat fraction (MRI-PDFF) based on liver magnetic resonance imaging was used to diagnose intrahepatic steatosis. Specifically, intrahepatic steatosis is defined as MRI-PDFF≥5%. For the identification of NAFLD individuals, ICD9 / 10 codes (such as 571.5, 571.8, 571.9, K74.0, K74.6, K75.8 and K76.0) or MRI-PDFF≥5% are first used. Meanwhile, individuals with other liver diseases were excluded by ICD9 or ICD10 codes (e.g., 571.1–4, 571.6, 572.0, 572.8, 573.3, 573.8–9, K70.0–4, K70.9, K71.0–2, K71.5–9, K72.0–1, K72.9, K73.0–2, K73.8–9, K74.1–5, K75.0, K75.2–4, K75.9, K76.1–3, K76.6–9, and K77.0).

[0068] In a cohort of healthy people undergoing physical examinations at a certain hospital, the definition of NAFLD is based on liver ultrasound examination. Ultrasound examination results can be used to divide individuals into normal control groups and different degrees of fatty liver, including mild, moderate and severe. Mild symptoms include slightly enhanced echogenicity of the liver parenchyma, moderate symptoms may be accompanied by echo attenuation of the right posterior lobe of the liver or unclear display of the diaphragm, and decreased echogenicity of the portal vein wall, while severe symptoms include echo attenuation of the liver parenchyma extending to the anterior lobe and blurred intrahepatic vascular walls.

[0069] It can be understood that by screening the first user information, clinical information and genetic information with high quality and relevance are selected.

[0070] It can be understood that, in some embodiments, after step S200, the method further includes: evaluating the performance of the non-alcoholic fatty liver disease diagnosis model according to a five-fold cross validation method.

[0071] Specifically, the performance of the non-alcoholic fatty liver disease diagnostic model was evaluated according to the five-fold cross-validation method, including: randomly dividing the screened risk clinical information and risk gene information into five equal parts, four of which were used as training sets and one as an internal validation set; in each iteration, the non-alcoholic fatty liver disease diagnostic model was trained according to the four training sets, and the non-alcoholic fatty liver disease diagnostic model was validated according to the internal validation set, and the performance index between the predicted result and the actual label of each non-alcoholic fatty liver disease diagnostic model was calculated until each training set was used as an internal validation set; the performance index in all iterations was obtained, and the model structure and parameters of the non-alcoholic fatty liver disease diagnostic model were adjusted according to the performance index in all iterations until the optimal non-alcoholic fatty liver disease diagnostic model was obtained.

[0072] In the embodiment of the present application, the number of training sets each time is 29192, and the number of internal test sets is 7298.

[0073] S300: Obtain clinical information and gene information in the training set and the internal test set.

[0074] It can be understood that in the embodiments of the present application, the clinical information includes age, gender, body mass index (BMI), blood glucose (Glu), triglyceride (TG), cholesterol (CHO), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), alanine aminotransferase (ALT) and aspartate aminotransferase (AST).

[0075] It is understandable that in the embodiments of the present application, the genetic information uses the genetic data of version 3 released by the UK Biobank in March 2018. After strict variant quality control screening, the genotype data was limited to 13.7 million single gene sites. The target genetic variation is a SNP that was identified as significantly associated with NAFLD (P<5×10-6) in the GWAS of the UK Biobank, or a SNP associated with NAFLD reported in previous studies. The genotype information of these SNPs was extracted as genetic features from the genotype dataset of the UK Biobank and the genotype dataset of a hospital health examination population cohort, and finally the genotype information of 646 SNPs that existed in both datasets was obtained.

[0076] S400: Obtain risk clinical information from clinical information.

[0077] In an embodiment of the present application, a method for constructing a diagnostic model for non-alcoholic fatty liver disease includes obtaining risk clinical information in clinical information. Specifically, the risk clinical information in the clinical information is mainly based on previous research results, and ten variables related to the risk of NAFLD are selected. The risk clinical information specifically includes age, gender, body mass index (Body massindex, BMI), blood glucose (blood glucose, Glu), triglyceride (triglyceride, TG), cholesterol (cholesterol, CHO), low-density lipoprotein cholesterol (low-density lipoprotein-cholesterol, LDL-C), high-density lipoprotein cholesterol (high-density lipoprotein-cholesterol, HDL-C), Alanine Aminotransferase (Alanine Aminotransferase, ALT) and Aspartate Aminotransferase (Aspartate Aminotransferase, AST), that is, the ten indicators counted in Table 1.

[0078] S500: Screening gene information according to the minimum absolute value shrinkage and selection operator regression to obtain risk gene information.

[0079] In an embodiment of the present application, a method for constructing a diagnostic model for non-alcoholic fatty liver disease screens gene information based on minimum absolute value shrinkage and selection operator regression to obtain risk gene information.

[0080] For details, please refer to Figure 4a and Figure 4b . Figure 4a This is a prediction performance diagram of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application. Figure 4a The mean square error (MSE) of different numbers of SNPs revealed in the training set using the lasso algorithm is shown, and the vertical dotted line marks the optimal parameter (i.e., lambda) value determined by the MSE minimum criterion. It can be understood that the minimum lambda value represents the lambda value with the minimum MSE obtained by 5-fold cross validation. At this time, the number of SNPs is 217, so there are a total of 217 risk genes used for modeling in this application.

[0081] It is understandable that in step S500, the gene information is screened using the minimum absolute value shrinkage and selection operator regression method to identify gene loci associated with the risk of NAFLD. This method can effectively reduce the dimension of the gene information and select the most predictive gene variables, thereby improving the accuracy and interpretability of the non-alcoholic fatty liver disease diagnostic model.

[0082] S600: Construct a non-alcoholic fatty liver disease diagnosis model, which includes a basic module and a meta-module.

[0083] Please also read Figure 3 , Figure 3 A block diagram of a non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0084] In an embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnosis model includes constructing a non-alcoholic fatty liver disease diagnosis model 30 , and the non-alcoholic fatty liver disease diagnosis model 30 includes a basic module 31 and a meta-module 32 .

[0085] In the embodiment of the present application, the basic module 31 and the meta-module 32 are connected in a cascade manner, and the basic module 31 includes a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression, and K nearest neighbor. These algorithms can process data of different types and features, and have different advantages and applicable scenarios. The meta-module 32 includes a logistic regression classifier, which is further predicted and classified based on the output of the basic module 31, thereby improving the overall performance and accuracy of the model.

[0086] It can be understood that the basic module 31 and the meta-module 32 are interconnected through a cascade connection. The basic module 31 is responsible for extracting features and performing preliminary classification, while the meta-module 32 performs deeper classification and prediction based on the basic module 31, comprehensively considering the output results of each algorithm, thereby constructing a more accurate and reliable non-alcoholic fatty liver disease diagnostic model 30.

[0087] S700: Training a diagnostic model for non-alcoholic fatty liver disease based on risk clinical information and risk gene information.

[0088] In an embodiment of the present application, the method for constructing a non-alcoholic fatty liver disease diagnosis model includes training a non-alcoholic fatty liver disease diagnosis model 30 based on risk clinical information and risk gene information.

[0089] Specifically, the training set was randomly divided into five different subsets of equal size, four of which were used for model training in each iteration, while the remaining subset was used for internal validation. This iterative process was performed a total of five times. In each iteration, seven basic classifiers were used to make predictions on the remaining subset to evaluate the performance of the basic classifiers relative to the true situation. At the same time, these prediction results were aggregated to form the input features of the meta-classifier. After completing five iterations, a complete input feature set was obtained, which contained the features generated by all base classifiers. This feature set was used to train the meta-classifier, and the XgBoost classifier was used to train the ensemble model. That is, the seven basic classifiers in the basic module independently predicted the clinical and genetic features in the training set, and then the fusion module used these prediction results to train the meta-classifier to predict NAFLD. Such a training method can improve the robustness and generalization ability of the model, enabling it to more accurately predict the incidence of non-alcoholic fatty liver disease.

[0090] S800: Obtaining a receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model. If the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use.

[0091] Please also read Figure 4c to Figure 4e , Figure 4c This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application. Figure 4d This is a predictive performance diagram of a non-alcoholic fatty liver disease diagnostic model provided in another embodiment of the present application. Figure 4e This is a prediction performance diagram of a non-alcoholic fatty liver disease diagnosis model provided in another embodiment of the present application. Specifically, Figure 4c ROC curves for predicting nonalcoholic fatty liver disease using the nonalcoholic fatty liver disease diagnostic model 30 and the traditional model (FIB-4) in the training set. Figure 4d ROC curves for the prediction of NAFLD by the NAFLD diagnostic model 30 and the traditional model (FIB-4) in the internal test set. Figure 4e ROC curves for the prediction of NAFLD in the external test set using the NAFLD diagnostic model 30 and the traditional model (FIB-4).

[0092] The non-alcoholic fatty liver disease diagnosis model 30 provided in the embodiment of the present application is superior to the traditional model in the diagnosis of non-alcoholic fatty liver disease and is applicable to different data sets. Through these figures, the performance of the non-alcoholic fatty liver disease diagnosis model 30 and the traditional model (FIB-4) on different data sets can be clearly compared, proving the superiority and versatility of the non-alcoholic fatty liver disease diagnosis model 30 provided in the embodiment of the present application.

[0093] Please also read Figure 5a . Figure 5a This is a prediction performance graph of the non-alcoholic fatty liver disease diagnosis model training set provided in an embodiment of the present application. Specifically, Figure 5a The performance of the non-alcoholic fatty liver disease diagnosis model 30 in the training set is compared with the traditional model (FIB-4). Figure 5a As shown, the AUC of the non-alcoholic fatty liver disease diagnostic model 30 in the training set for predicting the occurrence of severe liver disease was 0.851 (95%, CI: 0.811-0.891), while the AUC of FIB-4 was only 0.671 (95% CI: 0.609-0.732).

[0094] Please also read Figure 5b . Figure 5b This is a prediction performance graph of a non-alcoholic fatty liver disease diagnosis model training set provided by another embodiment of the present application. Specifically, Figure 5b Figure 3 shows the Kaplan-Meier curves of the incidence of severe liver disease according to the quartile grouping of the non-alcoholic fatty liver disease diagnostic model 30 in the internal test set. Figure 5b It can be seen that the increased incidence of non-alcoholic fatty liver disease is positively correlated with the incidence of severe liver disease.

[0095] In the embodiment of the present application, the threshold of the receiver operating characteristic curve is 0.8. It can be understood that the receiver operating characteristic curve is mainly used to evaluate the accuracy of the prediction model. The range of the area under the ROC curve (AUC) is 0.5-1.0, and AUC of 0.5 corresponds to random classification, 1.0 means completely accurate, 0.51-0.59 represents that the accuracy of the model is very poor, 0.60-0.69 represents poor accuracy, 0.70-0.79 represents medium accuracy, 0.80-0.89 represents good accuracy, and ≥0.90 represents very good accuracy. The results of the non-alcoholic fatty liver disease diagnostic model 30 provided in the present application are all above 0.8, indicating that the prediction accuracy of the model is high and has good diagnostic performance. Therefore, the non-alcoholic fatty liver disease diagnostic model 30 of the present application shows high accuracy and reliability in predicting non-alcoholic fatty liver disease, providing a useful tool for clinical practice.

[0096] In other embodiments, the non-alcoholic fatty liver disease diagnostic model is used to output outcome events, and the outcome events include adverse intrahepatic outcomes and adverse extrahepatic outcomes. Adverse intrahepatic outcomes include individuals with non-alcoholic fatty liver disease also having severe liver disease. Extrahepatic adverse outcomes include individuals with non-alcoholic fatty liver disease also having hypertension, type 2 diabetes, chronic kidney disease, coronary heart disease, heart failure, and atrial fibrillation or atrial flutter.

[0097] Specifically, the diagnosis of outcome events was determined based on ICD codes. The definition of diabetes mellitus type 2 (DmT2) could be determined based on one of the following criteria: self-reported DmT2 during the post-baseline follow-up period, hospital diagnosis of DmT2 occurring after the baseline assessment visit (determined by ICD code), current insulin therapy and / or use of oral hypoglycemic agents, serum glucose level ≥11.1 mmol / L (200 mg / dL), HbA1c ≥48 mmol / mol (equivalent to the traditional unit %: 6.5%). Chronic kidney disease (CKD) was defined as a diagnosis using ICD codes that occurred during the post-baseline follow-up period. The definition of coronary heart disease (CAD) included self-reported angina, myocardial infarction, or similar diagnoses during the post-baseline follow-up period, and ICD-coded diagnoses of CAD that occurred after the baseline assessment visit. The definition of heart failure (HF) covered self-reported heart failure, heart transplant surgery, or similar diagnoses during the post-baseline follow-up period, and CAD that occurred during the follow-up period. Atrial fibrillation or atrial flutter (AF) was defined as the occurrence of AF or atrial flutter according to ICD codes during follow-up. The integration of these definitions can help to more accurately identify and monitor complications in patients with NAFLD, thereby guiding clinical management and treatment decisions.

[0098] For details, please refer to Figure 6 and Figure 7a , Figure 7b , Figure 6 This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application. Figure 7a This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application. Figure 7b This is a graph of the prediction performance of adverse outcomes in the training set of the non-alcoholic fatty liver disease diagnosis model provided in one embodiment of the present application.

[0099] Specifically, Figure 7a Figure 3 ROC curves for predicting severe liver disease (SLD) using the nonalcoholic fatty liver disease diagnostic model 30 and the traditional model (FIB-4) in the external test set. Figure 7a As shown, in the results of applying the external test set to test the adverse outcomes, the non-alcoholic fatty liver disease diagnostic model 30 provided by the present application has an area under the ROC curve (AUC) higher than 0.8 (0.851) in the diagnosis of the cumulative incidence of severe liver disease (SLD), which is much higher than 0.671. This shows that the non-alcoholic fatty liver disease diagnostic model provided by the present application has higher accuracy and reliability in diagnosing severe liver disease, and has significant advantages over traditional models. The closer the area under the ROC curve is to 1, the better the diagnostic performance of the model.

[0100] Specifically, Figure 6 Figure 3 Kaplan-Meier curves of the incidence of hypertension, coronary heart disease, type 2 diabetes, heart failure, chronic kidney disease, and atrial fibrillation according to the quartile grouping of the nonalcoholic fatty liver disease diagnostic model 30 in the external test set. Figure 7b Figure 3 is the Kaplan-Meier curve of the incidence of severe liver disease according to the quartile grouping of the non-alcoholic fatty liver disease diagnostic model 30 in the external test set. Figure 6 and Figure 7b As shown, based on data from the external test set, an increased incidence of non-alcoholic fatty liver disease was positively correlated with the incidence of hypertension, coronary heart disease, type 2 diabetes, heart failure, chronic kidney disease, atrial fibrillation, and severe liver disease.

[0101] It is understandable that the application of the diagnostic model for non-alcoholic fatty liver disease is not limited to the simple diagnosis of non-alcoholic fatty liver disease, but extends to the prediction and analysis of outcome events. These outcome events include adverse intrahepatic outcomes and adverse extrahepatic outcomes, which aim to more comprehensively assess the health status and risk of disease in patients with NAFLD. Specifically, adverse intrahepatic outcomes refer to individuals who have been diagnosed with NAFLD but subsequently develop severe liver diseases such as cirrhosis and liver failure. Extrahepatic adverse outcomes are more extensive, including hypertension, type 2 diabetes, chronic kidney disease, coronary heart disease, heart failure, and a variety of complications such as atrial fibrillation or atrial flutter. By predicting and monitoring these outcome events, the diagnostic model for non-alcoholic fatty liver disease can not only provide important information about the disease progression and treatment effects of NAFLD patients, but also help doctors develop personalized treatment plans and preventive measures, thereby minimizing patients' health risks and improving their quality of life. That is, the application of the diagnostic model for non-alcoholic fatty liver disease in the management of NAFLD not only helps to detect disease risks early, but also promotes the implementation of more comprehensive health management and prevention strategies.

[0102] Figure 2 FIG. 1 is a schematic diagram of a non-alcoholic fatty liver disease diagnosis model construction system 10 provided in an embodiment of the present application. Figure 2 As shown, the non-alcoholic fatty liver disease diagnostic model construction system 10 includes at least the following parts: an information collection module 110 , an information screening module 120 , an information acquisition module 130 , a model construction module 140 and a model training module 150 .

[0103] In the embodiment of the present application, the information acquisition module 110 is used to obtain the first user information from the first data set. This information includes the patient's clinical data, biological sample information, and gene data. The information acquisition module 110 is used to collect this data to provide basic data for subsequent analysis and modeling.

[0104] In the embodiment of the present application, the information screening module 120 is used to screen the first user information to obtain clinical information and gene information in the first user information. Through screening, information closely related to the research purpose can be extracted from a large amount of data, thereby improving the efficiency and accuracy of modeling.

[0105] In the embodiment of the present application, the information acquisition module 130 is used to obtain clinical information and gene information in the training set and the internal test set, obtain risk clinical information in the clinical information, and screen the gene information according to the minimum absolute value shrinkage and selection operator regression to obtain risk gene information. This information is the key data required for building a non-alcoholic fatty liver disease diagnosis model, and the information acquisition module 130 ensures the completion of the data preparation work required for model establishment.

[0106] In an embodiment of the present application, the model building module 140 is used to construct a non-alcoholic fatty liver disease diagnosis model, which includes a basic module and a meta-module, wherein the basic module and the meta-module are connected in a cascade manner, the basic module includes a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression and K nearest neighbor, and the meta-module includes a logistic regression classifier.

[0107] In an embodiment of the present application, the model training module 150 is used to train the non-alcoholic fatty liver disease diagnosis model based on the risk clinical information and the risk gene information, and obtain the receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model. If the receiver operating characteristic curve is higher than the threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use. Through model training, the model can learn from the data and adjust the model parameters to improve the predictive performance of the model. After the training is completed, the module will also evaluate the performance of the model and generate a receiver operating characteristic curve (ROC curve) for further model evaluation and selection.

[0108] For details, please refer to Figure 1 And the corresponding description will not be repeated here.

[0109] Figure 8 The electronic device 20 provided by an embodiment of the present application is shown in FIG4 . As shown in FIG4 , the electronic device 20 at least includes the following parts: a processor 21 and a memory 22 .

[0110] In the embodiment of the present application, the memory 22 is used to store instructions executable by the processor 21, and the processor 21 is configured to implement the following when executing the instructions: Figure 1 The method for constructing a diagnostic model for non-alcoholic fatty liver disease is shown.

[0111] In an embodiment of the present application, a computer-readable storage medium includes instructions, and the instructions instruct a device to execute the method for constructing a non-alcoholic fatty liver disease diagnosis model as described in the first aspect. For example, the instructions instruct the device to execute Figure 1 The method for constructing a non-alcoholic fatty liver disease diagnosis model shown in steps S100 to S800.

[0112] The program operating in the electronic device 20 involved in one embodiment of the present application may be a program (a program that enables a computer to function) that controls a central processing unit (CPU) and the like to realize the functions of the above-mentioned embodiment involved in one scheme of the present invention. Then, the information processed by these devices is temporarily stored in a random access memory (RAM) during its processing, and then stored in various ROMs such as a read-only memory (Flash ROM) and a hard disk drive (HDD), and is read, modified, and written by the CPU as needed.

[0113] It should be noted that part of the electronic device 20 of the above embodiment may be implemented by a computer. In this case, a program for implementing the control function may be recorded in a computer-readable recording medium and implemented by reading the program recorded in the recording medium into a computer system and executing it.

[0114] It should be noted that the "computer system" mentioned here refers to a computer system built into the electronic device 20, and adopts a computer system including hardware such as an OS and peripheral devices. In addition, "computer-readable recording medium" refers to removable media such as a floppy disk, a magneto-optical disk, a ROM, a CD-ROM, and a storage device such as a hard disk built into the computer system.

[0115] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when sending programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memories inside computer systems that serve as servers or clients in this case. In addition, the above-mentioned program may be a program for realizing a part of the above-mentioned functions, or a program that can realize the above-mentioned functions by combining with a program already recorded in a computer system.

[0116] In addition, the electronic device 20 in the above-mentioned embodiment can also be implemented as a collection (device group) composed of multiple devices. Each device constituting the device group can have a part or all of the functions or functional blocks of the electronic device 20 in the above-mentioned embodiment. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device 20.

[0117] It can be understood that the method, system, device and medium for constructing a non-alcoholic fatty liver disease diagnostic model provided in the embodiments of the present application first screen the first user information to extract key clinical and genetic information related to NAFLD, and then apply minimum absolute value shrinkage and selection operator regression for gene screening to accurately capture risk gene information, and combine a variety of advanced machine learning algorithms to enhance the generalization and data processing capabilities of the model, which has diagnostic accuracy and reliability in the diagnosis of NAFLD compared to the existing technology.

[0118] It is understandable that the method, system, device and medium for constructing the diagnostic model of non-alcoholic fatty liver disease provided in the embodiments of the present application, first of all, can provide more comprehensive and in-depth data support for the early diagnosis of NAFLD by integrating large-scale clinical and genotype data from the UK Biobank and a hospital. Secondly, using advanced machine learning algorithms and gene information screening technology, the model can effectively identify and analyze potential NAFLD risk factors, and provide clinicians with more accurate diagnostic results and personalized treatment recommendations. In addition, through the verification of internal and external test sets, the robustness and reliability of the model have been fully verified, providing a solid foundation for its wide application in clinical practice. In summary, the NAFLD diagnostic model provided by this application has important beneficial effects in improving diagnostic accuracy, optimizing treatment options and promoting application in clinical practice, and is expected to provide reliable support for alleviating the pain of NAFLD patients and promoting health management.

[0119] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present application and are not intended to be limiting of the present application. As long as they are within the spirit and scope of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. A method for constructing a diagnostic model for non-alcoholic fatty liver disease, characterized in that: include: Obtaining user information from a first data set; Screening the user information from the first data set to obtain clinical information and gene information from the user information from the first data set; screening the gene information according to least absolute value shrinkage and selection operator regression to obtain risk gene information; The screened clinical information and risk gene information are randomly divided into five equal parts, four of which are used as training sets and one is used as an internal validation set; Obtaining clinical information and gene information in the training set and the internal validation set; Acquire user information from the second data set, and filter the user information from the second data set to acquire an external test set; The internal validation set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnostic model on processing training data, and the external test set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnostic model on processing unseen data; obtaining risk clinical information from the clinical information; Constructing a non-alcoholic fatty liver disease diagnosis model, the non-alcoholic fatty liver disease diagnosis model includes a basic module and a meta-module, wherein the basic module and the meta-module are connected in a cascade manner, the basic module includes a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression and K nearest neighbor, and the meta-module includes a logistic regression classifier; Training the non-alcoholic fatty liver disease diagnostic model according to the risk clinical information and the risk gene information; Evaluating the performance of the non-alcoholic fatty liver disease diagnostic model according to a five-fold cross-validation method; The five-fold cross-validation method comprises: in each iteration, training the non-alcoholic fatty liver disease diagnosis model according to the four training sets, validating the non-alcoholic fatty liver disease diagnosis model according to the internal validation set, and calculating the performance index between the prediction result of the non-alcoholic fatty liver disease diagnosis model and the actual label each time, until each of the training sets is used as the internal validation set; Obtain performance indicators in all iterations, and adjust the model structure and parameters of the non-alcoholic fatty liver disease diagnosis model according to the performance indicators in all iterations until the optimal non-alcoholic fatty liver disease diagnosis model is obtained; obtain the receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model, and if the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use.

2. The method for constructing a diagnostic model for non-alcoholic fatty liver disease according to claim 1, characterized in that: The method for constructing a non-alcoholic fatty liver disease diagnostic model further comprises: The non-alcoholic fatty liver disease diagnostic model is used to output outcome events, and the outcome events include adverse intrahepatic outcomes and adverse extrahepatic outcomes; The intrahepatic adverse outcomes include individuals with NAFLD also developing severe liver disease; Extrahepatic adverse outcomes include hypertension, type 2 diabetes, chronic kidney disease, coronary heart disease, heart failure, and atrial fibrillation or flutter in individuals with nonalcoholic fatty liver disease.

3. The method for constructing a diagnostic model for non-alcoholic fatty liver disease according to claim 1, characterized in that: The method for constructing a non-alcoholic fatty liver disease diagnostic model further comprises: Grid search and Bayesian optimization were used to determine the optimal hyperparameters of the non-alcoholic fatty liver disease diagnostic model.

4. A system for constructing a diagnostic model for non-alcoholic fatty liver disease, applied to the method for constructing a diagnostic model for non-alcoholic fatty liver disease as claimed in any one of claims 1 to 3, characterized in that: include: An information collection module, used to obtain user information from the first data set; An information screening module, used for screening the user information from the first data set to obtain clinical information and gene information from the user information from the first data set; The screened clinical information and risk gene information are randomly divided into five equal parts, four of which are used as training sets and one is used as an internal validation set; An information acquisition module, used to acquire clinical information and gene information from the training set and the internal validation set, acquire user information from a second data set, and screen the user information from the second data set to acquire an external test set; wherein the internal validation set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnostic model on processing training data, and the external test set is used to evaluate the performance of the non-alcoholic fatty liver disease diagnostic model on processing unseen data; acquire risk clinical information from the clinical information, and screen the gene information according to minimum absolute value shrinkage and selection operator regression to acquire risk gene information; A model building module, used for building a non-alcoholic fatty liver disease diagnosis model, wherein the non-alcoholic fatty liver disease diagnosis model includes a basic module and a meta-module, wherein the basic module and the meta-module are connected in a cascade manner, the basic module includes a lightweight gradient boosting machine, classic extreme gradient boosting, category boosting, random forest, adaptive boosting, logistic regression and K nearest neighbor, and the meta-module includes a logistic regression classifier; A model training module is used to train the non-alcoholic fatty liver disease diagnosis model according to the risk gene information, and evaluate the performance of the non-alcoholic fatty liver disease diagnosis model according to the five-fold cross-validation method; the five-fold cross-validation method includes: in each iteration, the non-alcoholic fatty liver disease diagnosis model is trained according to the four training sets, and the non-alcoholic fatty liver disease diagnosis model is verified according to the internal validation set, and the performance index between the predicted result of the non-alcoholic fatty liver disease diagnosis model and the actual label is calculated each time until each training set is used as the internal validation set; the performance index in all iterations is obtained, and the model structure and parameters of the non-alcoholic fatty liver disease diagnosis model are adjusted according to the performance index in all iterations until the optimal non-alcoholic fatty liver disease diagnosis model is obtained; the receiver operating characteristic curve of the non-alcoholic fatty liver disease diagnosis model is obtained, and if the receiver operating characteristic curve is higher than a threshold, the non-alcoholic fatty liver disease diagnosis model can be put into use.

5. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the method for constructing a non-alcoholic fatty liver disease diagnosis model as described in any one of claims 1 to 3 when executing the instructions.

6. A computer-readable storage medium, characterized in that: The method comprises instructions for instructing a device to execute the method for constructing a non-alcoholic fatty liver disease diagnosis model as described in any one of claims 1 to 3.

Images