Method and system for risk assessment of metabolic-related fatty liver disease, and electronic device

Abstract

The application provides a risk assessment method and system for metabolic associated fatty liver disease, and an electronic device, comprising: acquiring a plurality of initial anthropometric parameters and initial metabolic parameters of a target object, determining a plurality of candidate indexes based on the plurality of initial anthropometric parameters and initial metabolic parameters; screening a target index combination from the plurality of candidate indexes; determining an evaluation index for evaluating the risk of metabolic associated fatty liver disease based on the target index combination, and evaluating the risk of metabolic associated fatty liver disease of the target object according to the evaluation index. The technical problem of inaccurate evaluation and poor population universality caused by relying on a single or a few non-optimized indexes in the prior art is solved, the overall accuracy of metabolic associated fatty liver disease risk identification is significantly improved, the precise locking of high-risk groups is realized, and the prediction accuracy of MASLD related diseases is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology for metabolic-associated fatty liver disease, specifically to a risk assessment method, system, and electronic device for metabolic-associated fatty liver disease. Background Technology

[0002] Metabolic dysfunction-associated steatotic liver disease (MASLD) is a common chronic liver disease caused by hepatic steatosis accompanied by metabolic disorders. Early identification and risk assessment are crucial for delaying disease progression and improving patient prognosis. However, current routine indicators used to assess obesity and metabolic risk are not very effective in diagnosing MASLD. For example, while Body Mass Index (BMI) is a universally accepted measure of obesity, it fails to reflect differences in fat distribution, only indicating the relationship between total weight and height. It cannot distinguish between visceral adipose tissue, which is highly lipotoxic and pro-inflammatory, and relatively benign subcutaneous adipose tissue. Furthermore, BMI cannot identify sarcopenic obesity, a common phenomenon in liver disease patients where muscle loss masks fat gain. This means that while a patient's BMI may appear normal, their internal metabolic environment is actually severely disrupted, making it easy to miss or misdiagnose based solely on BMI. Waist circumference (WC), while reflecting abdominal obesity, has varying threshold standards across different populations, and a single indicator like waist circumference is insufficient to comprehensively assess systemic metabolic risk. Novel metabolic indices such as the triglyceride-glycemic index (TyG) are correlated with fatty liver, but their diagnostic efficacy is limited when used alone. In summary, existing technologies suffer from low diagnostic efficiency, inconsistent standards, and limitations of single indicators, making it difficult to meet the needs of clinical screening and diagnosis of MASLD.

[0003] Therefore, a large number of missed diagnoses and misdiagnoses have emerged in clinical practice. On the one hand, many patients with "lean MASLD" (BMI<25) who have normal weight but central obesity and insulin resistance are missed in routine BMI screening. On the other hand, some people with "metabolic healthy obesity" are over-medicalized, but there is currently a lack of a unified non-invasive obesity-related indicator that has been validated on a large scale to accurately identify MASLD.

[0004] In summary, there is currently a lack of a non-invasive risk assessment tool that can efficiently integrate multi-dimensional anthropometric and metabolic information and has been optimized through large-scale data-driven construction to address the problems of low accuracy and poor universality in MASLD clinical screening.

[0005] Therefore, the existing technology still needs further development. Summary of the Invention

[0006] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide a risk assessment method, system, and electronic device for metabolic-related fatty liver disease, so as to solve the problems existing in the prior art.

[0007] To achieve the above-mentioned technical objectives, according to a first aspect of the present invention, the present invention provides a risk assessment method for metabolic-related fatty liver disease, comprising: S100: Obtain multiple initial anthropometric parameters and initial metabolic parameters of the target object, and determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; S200. Select a target indicator combination from the plurality of candidate indicators; S300. Based on the target indicator combination, determine the assessment index for evaluating the risk of metabolic-related fatty liver disease, and conduct a risk assessment of metabolic-related fatty liver disease on the target subject according to the assessment index.

[0008] Specifically, the method for selecting a combination of target indicators from the plurality of candidate indicators includes: The LASSO regression algorithm is used to perform feature selection on the multiple candidate indicators in order to screen out the target indicator combination and determine the weight coefficients corresponding to each target indicator.

[0009] Specifically, the method for selecting the combination of target indicators and determining the weight coefficients corresponding to each target indicator includes: Using the multiple candidate indicators as independent variables and the prevalence of metabolic-related fatty liver disease as the dependent variable, a LASSO regression model was constructed. By adjusting the penalty parameter of the LASSO regression model, a subset of target indicators is selected from the multiple candidate indicators; Based on the regression coefficients of each indicator in the LASSO regression model within the target indicator subset, the weight coefficients corresponding to each target indicator are determined.

[0010] Specifically, the process of selecting the target indicator combination also includes: Multicollinearity is judged on the target indicator combination obtained by the LASSO regression algorithm. Based on the judgment result, derivative indicators that are collinear with at least one other target indicator are removed from the target indicator combination obtained by the screening. The derived index is a composite index calculated from at least two initial anthropometric parameters or initial metabolic parameters.

[0011] Specifically, the combination of target indicators includes: Waist circumference, waist-to-height ratio, waist-to-hip ratio, triglyceride-glucose index, conic index, and lipid accumulation products.

[0012] Specifically, the method for determining the assessment index for evaluating the risk of metabolic-related fatty liver disease based on the target indicator combination includes: The evaluation index is obtained by taking each index in the target index combination after multicollinearity processing as input variables, and taking the weight coefficients corresponding to each target index determined by the LASSO regression algorithm as weight coefficients of each input variable, and performing a weighted summation operation.

[0013] Specifically, the obtained evaluation index is represented as follows: MBRI=k1×WC+k2×WHtR+k3×WHpR+k4×TyG−k5×LAP+k6×CI; Wherein, MBRI represents the assessment index, WC represents waist circumference, k1 represents the weighting coefficient of waist circumference, WHtR represents waist-to-height ratio, k2 represents the weighting coefficient of waist-to-height ratio, WHpR represents waist-to-hip ratio, k3 represents the weighting coefficient of waist-to-hip ratio, TyG represents the triglyceride-glucose index, k4 represents the weighting coefficient of triglyceride-glucose index, LAP represents lipid accumulation products, k5 represents the weighting coefficient of lipid accumulation products, CI represents the conic index, and k6 represents the weighting coefficient of conic index.

[0014] Specifically, the risk assessment of metabolic-related fatty liver disease for the target subject based on the assessment index includes: The assessment index is compared with a preset threshold, and the risk assessment result of the target object regarding metabolic-related fatty liver disease is output based on the comparison result.

[0015] According to a second aspect of the present invention, a risk assessment system for metabolic-related fatty liver disease is provided, comprising: Acquisition module: used to acquire multiple initial anthropometric parameters and initial metabolic parameters of the target object, and to determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; Indicator filtering module: used to filter out a target indicator combination from the multiple candidate indicators; Risk assessment module: used to determine an assessment index for assessing the risk of metabolic-associated fatty liver disease based on the target indicator combination, and to conduct a risk assessment of metabolic-associated fatty liver disease on the target subject according to the assessment index.

[0016] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, which, when executed by the processor, implement the above-described risk assessment method for metabolic-related fatty liver disease.

[0017] Beneficial effects: This invention provides a risk assessment method and system for metabolic-associated fatty liver disease (MASLD), which solves the technical problems of inaccurate assessment and poor population universality caused by relying on single or a few non-optimized indicators in the prior art. By automatically screening and integrating a risk assessment index that can comprehensively reflect an individual's fat distribution characteristics and metabolic disorder status from multi-dimensional anthropometric and metabolic parameters, the invention significantly improves the overall accuracy of risk identification for MASLD, achieves precise identification of high-risk groups, and greatly improves the prediction accuracy of MASLD-related diseases. Attached Figure Description

[0018] Figure 1 This is a flowchart of a risk assessment method for metabolic-related fatty liver disease provided in a specific embodiment of the present invention; Figure 2 This is a schematic diagram of the system composition of the risk assessment system for metabolic-related fatty liver disease provided in a specific embodiment of the present invention; Figure 3 This is a diagnostic ROC curve for MASLD in the validation set provided in a specific embodiment of the present invention; Figure 4 This is a ROC curve for NAFLD diagnosis in a validation set provided in a specific embodiment of the present invention; Figure 5 This is a specific embodiment of the present invention providing an at-risk MASH diagnostic ROC curve in the validation set; Figure 6 This is a diagnostic ROC curve for severe MASLD in the validation set provided in a specific embodiment of the present invention. Figure 7 This is a decision curve analysis diagram of MBRI and a single indicator under MASLD provided in a specific embodiment of the present invention. Figure 8 This is a decision curve analysis diagram of MBRI and a single indicator under NAFLD provided in a specific embodiment of the present invention. Figure 9 This is a decision curve analysis diagram of MBRI and a single indicator under at-risk MASH provided in a specific embodiment of the present invention; Figure 10 This is a decision curve analysis diagram of MBRI and a single indicator under severe MASLD provided in a specific embodiment of the present invention. Figure 11 This is a diagnostic ROC curve of MASLD in the test set provided in a specific embodiment of the present invention; Figure 12 This is a diagnostic ROC curve of NAFLD in the test set provided in a specific embodiment of the present invention; Figure 13 This is a diagnostic ROC curve of at-risk MASH in the test set provided in a specific embodiment of the present invention; Figure 14 This is a diagnostic ROC curve of severe MASLD in the test set provided in a specific embodiment of the present invention; Figure 15 This is the SHAP beeswarm plot of the GLM model provided in a specific embodiment of the present invention; Figure 16 This is the SHAP beeswarm plot of the GBM model provided in a specific embodiment of the present invention; Figure 17 This is the SHAP beeswarm diagram of the NNET model provided in a specific embodiment of the present invention; Figure 18 This is a bar chart of the average absolute SHAP value of the GLM model provided in a specific embodiment of the present invention; Figure 19 This is a bar chart of the average absolute SHAP value of the GBM model provided in a specific embodiment of the present invention; Figure 20 This is a bar chart of the average absolute SHAP value of the NNET model provided in a specific embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Other similar embodiments obtained by those skilled in the art based on the embodiments in this application without creative effort should all fall within the scope of protection of this application. Furthermore, directional terms mentioned in the following embodiments, such as "up," "down," "left," and "right," are only for reference to the directions in the accompanying drawings; therefore, the directional terms used are for illustrative purposes and not for limiting the invention.

[0020] The present invention aims to construct a non-invasive MASLD Body Risk Index (MBRI) for metabolic-associated fatty liver disease (MASLD), which is an assessment index. The MBRI is based on multiple anthropometric indicators and metabolic-related indicators. It uses the LASSO regression algorithm for feature selection and parameter estimation to form a linear weighted index to quantify an individual's risk level of developing MASLD.

[0021] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0022] Example 1 Please see Figure 1This embodiment provides a risk assessment method for metabolic-associated fatty liver disease, including: acquiring multiple initial anthropometric parameters and initial metabolic parameters of a target subject; determining multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; selecting a target indicator combination from the multiple candidate indicators; determining an assessment index for assessing the risk of metabolic-associated fatty liver disease based on the target indicator combination; and performing a risk assessment of metabolic-associated fatty liver disease on the target subject according to the assessment index.

[0023] Understandably, by adopting the above technical solution, the technical problems of inaccurate assessment and poor population universality caused by relying on a single or a few non-optimized indicators in the existing technology have been solved. By automatically screening and integrating a risk assessment index that can comprehensively reflect an individual's fat distribution characteristics and metabolic disorder status from multi-dimensional anthropometric and metabolic parameters, the overall accuracy of risk identification of metabolic-related fatty liver disease has been significantly improved, and the precise identification of high-risk groups has been achieved, which has greatly improved the prediction accuracy of MASLD-related diseases.

[0024] See Figure 1 The specific implementation process of the risk assessment method for metabolic-related fatty liver disease in this embodiment is as follows: S100: Obtain multiple initial anthropometric parameters and initial metabolic parameters of the target object, and determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; It should be noted that initial anthropometric parameters refer to data that can be directly measured through standardized physical examinations. These typically include, but are not limited to, height (Ht), weight (Wt), waist circumference (WC), and hip circumference (Hip). These parameters reflect the basic body shape and fat distribution of the target individual. They can be obtained through clinical measurement, automatic reading by medical examination equipment, or user self-entry, and must comply with medical measurement standards. Initial metabolic parameters refer to core biochemical indicators that reflect the body's metabolic state, obtained through laboratory blood tests. These include, at least, fasting serum triglyceride concentration (TG) and plasma glucose concentration (Glu). These data are usually obtained from the target individual's fasting blood test report. Based on the initial parameters obtained above, a series of derived indicators can be calculated using predefined mathematical formulas or rules to form a set of candidate indicators for feature screening. These candidate indicators aim to quantify obesity type, body fat distribution, and degree of metabolic disorder from different perspectives. The determined candidate indicators include various types directly or indirectly calculated from the initial parameters, such as traditional obesity indicators, such as body mass index (BMI); body shape distribution indicators, such as waist-to-height ratio (WHtR), conic index (CI); metabolic composite index, such as triglyceride-glucose index (TyG); and other composite indicators, etc., which lay the foundation for subsequent screening of the optimal prediction combination using advanced statistical methods.

[0025] Preferably, this embodiment directly utilizes large-sample epidemiological data from the US NHANES database, incorporating 17 candidate anthropometric indicators, including traditional obesity indicators and newly proposed metabolic indices. These candidate variables cover various aspects of obesity degree, body shape distribution, and basal metabolism, such as BMI, waist circumference (WC), hip circumference (HIP), waist-to-height ratio (WHtR), waist-to-hip ratio (WHpR), conic index (CI), body shape index (ABSI), lipid accumulation products (LAP), visceral fat index (VAI), body surface area (BSA), body fat index (BAI), body roundness index (BRI), and various composite indicators based on triglycerides and blood glucose (TyG and its combination with BMI / waist circumference / height—TyG-BMI, TyG-WHtR, TyG-WHpR).

[0026] S200. Select a target indicator combination from the plurality of candidate indicators; In this embodiment, the method for selecting a target indicator combination from the plurality of candidate indicators includes: The LASSO regression algorithm is used to perform feature selection on the multiple candidate indicators in order to screen out the target indicator combination and determine the weight coefficients corresponding to each target indicator. Using the multiple candidate indicators as independent variables and the prevalence status of metabolic-related fatty liver disease as the dependent variable, a LASSO regression model is constructed. By adjusting the penalty parameter of the LASSO regression model, a subset of target indicators is selected from the multiple candidate indicators. Based on the regression coefficients of each indicator in the LASSO regression model within the subset of target indicators, the weight coefficients corresponding to each target indicator are determined.

[0027] Specifically, in the process of screening the target indicator combination, the target indicator combination obtained by the LASSO regression algorithm is subjected to multicollinearity judgment. Based on the judgment result, derivative indicators that are collinear with at least one other target indicator are removed from the screened target indicator combination. The derived index is a composite index calculated from at least two initial anthropometric parameters or initial metabolic parameters.

[0028] Furthermore, the combination of target indicators obtained through the above screening process is as follows: Waist circumference, waist-to-height ratio, waist-to-hip ratio, triglyceride-glucose index, conic index, and lipid accumulation products.

[0029] In a preferred embodiment, to select the variable most predictive of MASLD from the candidate indicators, this embodiment preferably uses LASSO (Least Absolute Shrinkage and Selection). LASSO regression uses a regression algorithm for feature selection. While introducing regularization, it compresses unimportant coefficients, enabling the model to automatically filter out redundant or severely collinear indicators, thus determining a concise and informative combination of variables. A unique advantage of LASSO regression is its ability to compress unimportant variable coefficients to zero, achieving automated feature selection. By performing LASSO screening on the training set data, and using MASLD as the outcome variable, the optimal feature set was obtained when the penalty function parameter was 10.76. Seven key variables were selected: waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHpR), triglyceride-glucose index (TyG), conic index (CI), lipid accumulation products (LAP), and triglyceride-glucose waist-to-hip ratio (TyG-WHpR). These variables collectively form the basis of the MBRI index. The regression coefficients corresponding to the above indicators are shown in Table 1. Table 1 Regression coefficients of each target indicator It should be further noted that during model development, collinearity testing was required for the aforementioned target indicators. Although the derived variable triglycerides, glucose, waist-to-hip ratio (TyG-WHpR) was initially included, it exhibited high collinearity with its constituent elements. Variance inflation factor (VIF) testing revealed that its VIF value far exceeded the threshold, indicating severe multicollinearity. To ensure model robustness, this highly collinear term needed to be discarded, retaining only the main basic indicators. Through this process, the VIF of all selected variables decreased to below 10, indicating collinearity. The problem was brought under control, and ultimately, waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHpR), triglyceride-glucose index (TyG), conic index (CI), and lipid accumulation products (LAP) were established as the six core components of the MBRI. They were linearly weighted according to their LASSO regression coefficients (weighting coefficients) to form a new evaluation index, MBRI. Compared with the traditional stepwise regression method, the LASSO regression algorithm can objectively and efficiently determine the optimal set of variables and avoid multicollinearity interference.

[0030] S300. Based on the target indicator combination, determine the assessment index for evaluating the risk of metabolic-related fatty liver disease, and conduct a risk assessment of metabolic-related fatty liver disease on the target subject according to the assessment index.

[0031] Specifically, a method for determining assessment indices for evaluating the risk of metabolic-related fatty liver disease based on the aforementioned combination of target indicators includes: The evaluation index is obtained by taking each index in the target index combination after multicollinearity processing as input variables, and taking the weight coefficients corresponding to each target index determined by the LASSO regression algorithm as weight coefficients of each input variable, and performing a weighted summation operation. Specifically, the evaluation index is expressed as follows: MBRI=0.0073×WC+4.2025×WHtR+2.1155×WHpR+0.6214×TyG−0.0023×LAP+0.1504×CI; Among them, MBRI represents the assessment index, WC represents waist circumference, WHtR represents waist-to-height ratio, WHpR represents waist-to-hip ratio, TyG represents triglyceride-glucose index, LAP represents lipid accumulation products, and CI represents conic index.

[0032] It should be noted that after obtaining the key variables, i.e., the combination of target indicators, the above target indicators are linearly combined according to their LASSO regression coefficients to define a new assessment index, MBRI. According to the above technical solution, by integrating multiple complementary indicators, MBRI can characterize the patient's metabolic disorders and fatty liver risk from different perspectives, compensating for the lack of information dimension of a single indicator. This embodiment introduces regularized feature screening to avoid collinear interference and forms an interpretable linear combination index for convenient clinical application. Compared with existing technical solutions, the key improvement of MBRI lies in the integration of multi-source information and data-driven optimization, which is expected to significantly improve the accuracy and stability of non-invasive diagnosis of MASLD.

[0033] In a preferred embodiment, the LASSO regression algorithm is used to perform feature selection on the multiple candidate indicators to screen out the target indicator combination and determine the weight coefficients corresponding to each target indicator, thereby obtaining the non-invasive evaluation index of MASLD. The process is as follows: (1) LASSO regression calculation In this embodiment, the presence or absence of MASLD (1 for yes, 0 for no) is used as the dependent variable, and the above-mentioned multiple anthropometric and metabolic indicators are used as independent variables. The model is performed using logistic regression with L1 regularization (i.e., LASSO logistic regression). Let the first The vector of independent variables for the subjects is ; in, This represents the number of candidate indicators (17 in this example). ∈{0,1} is the MASLD state indicator variable (1 indicates the existence of MASLD, 0 indicates its non-existence), and the logistic regression model is in the form of: ; in, Indicates that in the known number of... The features (or covariates) of each sample are The conditional probability of taking the value 1; LASSO constrains the regression coefficients and achieves variable selection by adding an L1 regularization term to the log-likelihood function. The optimization problem it solves is: ; in, For sample size, ≥0 is the penalty parameter, when When it increases, more It is compressed to 0, thereby enabling automatic variable filtering. For the intercept term, For the first The regression coefficients of the independent variables.

[0034] (2) Data sources and candidate variables This embodiment uses data obtained from the National Health and Nutrition Examination Survey (NHANES) database. First, the data was cleaned, samples with missing key variables were removed, and the target study population was selected as candidate independent variables based on inclusion and exclusion criteria. This embodiment includes 17 anthropometric and metabolic-related indicators, covering multiple dimensions such as obesity level, fat distribution, and metabolic load, including: 1) Traditional obesity and body shape indicators: Body Mass Index (BMI), Waist Circumference (WC), Hip Circumference (Hip), Waist-to-Height Ratio (WHtR), Waist-to-Hip Ratio (WHpR), Cone Index (CI), Body Shape Index (ABSI), Body Surface Area (BSA), Body Fat Index (BAI), and Body Circumference Index (BRI). 2) Metabolic and lipid-related indicators: Lipid accumulation products (LAP), visceral fat index (VAI), triglyceride-glucose index (TyG) and their combined indicators (TyG-BMI, TyG-WHtR, TyG-WHpR). (3) The specific calculation process of LASSO regression (implemented based on R language glmnet) This implementation uses the glmnet package in R language to perform LASSO logistic regression modeling and variable selection. The specific steps are as follows: 1) Data partitioning The NHANES samples that meet the criteria are randomly divided into training and validation sets in a ratio of approximately 7:3. 2) Model Setup In the training set, MASLD (0 / 1) is used as the dependent variable, and 17 candidate metrics are used as independent variables. The cv.glmnet function is called to perform LASSO modeling with cross-validation. The random seed is fixed by set.seed(2020). The main parameters are set as follows: family = binomial (binomial distribution, applicable to binary outcomes); Alpha = 1 (indicates the use of L1 regularization, i.e., LASSO); nfolds = 10 (using 10-fold cross-validation); Use the default settings for the remaining parameters.

[0035] 3) Selection of penalty parameter λ cv.glmnet runs 10-fold cross-validation on a series of candidate λ values ​​and calculates the mean cross-validation deviation (deviance) and its standard error for each λ. To balance model fit and simplicity, this embodiment adopts the "1-SE rule", which means that near the minimum cross-validation error, the solution with an error not exceeding one standard error of the minimum value and a larger λ value is selected as the final penalty parameter. The variable corresponding to this penalty parameter in the glmnet output is lambda.1se, which is denoted as λ_1se in this embodiment, and its logarithm is approximately 10.76. 4) Extraction of regression coefficients After determining λ_1se, the model coefficients are extracted by coef(cvfit, s = “lambda.1se”), which yields a sparse coefficient vector including the intercept term. Variables with coefficients of 0 are considered to be eliminated by LASSO, and variables with non-zero coefficients are the key features selected by LASSO. (4) Variable selection results and collinearity treatment Under the λ_1se condition, the LASSO logistic regression model identified seven variables with non-zero coefficients: waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHpR), triglyceride-glucose index (TyG), triglyceride-glucose waist-to-hip ratio (TyG-WHpR), conic index (CI), and lipid accumulation products (LAP). Their corresponding regression coefficients (rounded to four decimal places) are as follows: Intercept: -12.3640; WC: 0.0073; WHtR: 4.2025; WHpR: 2.1155; TyG: 0.6214; TyG-WHpR: 0.0571; CI: 0.1504; LAP: -0.0023; To further improve model stability and reduce multicollinearity, this embodiment performed variance inflation factor (VIF) analysis on the above seven variables. The results showed that TyG-WHpR exhibited strong collinearity with other variables, and the VIF value was significantly high. Therefore, this embodiment prioritized the removal of the TyG-WHpR variable. After removing TyG-WHpR, the VIF of the remaining six variables (WC, WHtR, WHpR, TyG, CI, and LAP) was recalculated. The results showed that the VIF values ​​of the above variables all decreased to below 10, and the multicollinearity problem was effectively controlled. Finally, this embodiment determined waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHpR), triglyceride-glucose index (TyG), conic index (CI), and lipid accumulation products (LAP) as the six core variables for constructing the MASLD Body Risk Index (MBRI).

[0036] (5) The final calculation formula of MBRI Based on the above, this embodiment linearly weights and combines the above six core variables according to their LASSO regression coefficients to construct the MASLD Body Risk Index (MBRI), whose linear prediction expression is as follows: MBRI = -12.3640 + 0.0073×WC + 4.2025×WHtR + 2.1155×WHpR + 0.6214×TyG + 0.1504×CI - 0.0023×LAP; In practical applications, an appropriate MBRI cutoff value can be set based on the ROC curve analysis results to distinguish between high-risk and low-risk individuals, and its diagnostic efficacy can be compared with traditional indicators (such as BMI, TyG, LAP, etc.), thereby achieving non-invasive risk stratification and early identification of MASLD.

[0037] In some specific embodiments, a method for assessing the risk of metabolic-related fatty liver disease in the target subject based on the assessment index includes: The assessment index is compared with a preset threshold, and the risk assessment result of the target object regarding metabolic-related fatty liver disease is output based on the comparison result.

[0038] It should be noted that the assessment index is a continuous numerical variable, and its value is positively correlated with the potential risk of the target subject developing metabolic-associated fatty liver disease (MASLD). In order to transform this continuous risk measure into an intuitive and actionable clinical judgment, it needs to be compared with one or more pre-set discrimination thresholds.

[0039] Specifically, the preset threshold can be determined based on large-scale, representative population study data, optimized through statistical methods to ensure the scientific validity and robustness of its discrimination power. Methods for determining the threshold include, but are not limited to: (1) Receiver operating characteristic curve analysis: In the model development or validation cohort, the ROC curve of the evaluation index for diagnosing MASLD was plotted, and the index value that maximizes the Youden index was selected as the optimal diagnostic threshold. (2) Clinical risk stratification needs: Different thresholds can be set according to different screening or management goals (such as prioritizing initial screening sensitivity or diagnostic specificity). For example, a lower screening threshold can be set to identify potential high-risk groups to the maximum extent, and a higher diagnostic reference threshold can be set to indicate highly suspected cases.

[0040] (3) Clinical validation studies: In an independent validation cohort, the positive predictive value, negative predictive value, and net clinical benefit corresponding to different thresholds are evaluated to ultimately determine the recommended thresholds applicable to the target population. Threshold information can be pre-stored or configured in the risk assessment system as part of the model parameters.

[0041] Furthermore, the actual assessment index calculated for the target object is compared with a preset threshold. Based on the comparison result, the MASLD risk of the target object is classified into a predefined risk level. For example, in a binary classification embodiment: if the assessment index is higher than or equal to the preset high-risk threshold, the target object is determined to be of high MASLD risk; if the assessment index is lower than the threshold, it is determined to be of low MASLD risk.

[0042] It should be noted that in more refined tiered management implementations, multiple thresholds can be used to achieve multi-level classification (e.g., low risk, medium risk, high risk), recommending differentiated follow-up management strategies for individuals with different risk levels. Outputting risk assessment results means generating and presenting a clear judgment conclusion based on the above classification logic. The output can be a direct classification label, displaying MASLD high risk or MASLD low risk in reports, user interfaces, or system messages. The output can also be a quantified risk probability; in some models, the assessment index can be directly or through a transformation function mapped to a disease risk probability (e.g., 0-100%), outputting this probability value, for example, a MASLD risk probability of 65%. The output can also be a structured report, whose results can be integrated into a more complete health assessment report, including the risk level, corresponding clinical recommendations (e.g., recommendations for liver ultrasound, lifestyle interventions, and regular follow-up), and key input values ​​used to calculate the index.

[0043] It is understandable that the above technical solutions, from evidence-based threshold setting to clear comparison and classification rules, and then to specific and readable result output formats, significantly improve the integrity and accuracy of the risk assessment process, and can be more reliably applied to clinical screening or health management scenarios, further expanding the application scenarios of the present invention.

[0044] Furthermore, the working principle of the present invention will be illustrated below with specific examples: Step 1: Data Preparation This example uses data from the NHANES survey from 2017 to 2020. The survey initially included 15,560 participants. After excluding 5,857 people aged 18 and under and 897 people diagnosed with other liver diseases (such as hepatitis B, hepatitis C, autoimmune hepatitis, or liver cancer), 8,806 participants were obtained. Further exclusions included 1,659 participants lacking BMI, WC, CAP, and LSM data, 373 participants with fasting time less than 3 hours, 624 participants with unreliable VCTE measurements (interquartile / median ratio ≥30%), and 817 participants with other missing key data. Therefore, the final analysis included 5,297 participants, and the dataset was randomly divided into a training set (n=3707) and an independent test set (n=1590) in a 7:3 ratio. Step 2: Definition criteria for clinical outcomes This example uses vibration-controlled transient elastography as a reference standard for liver lesions, with the following specific definitions: Hepatic steatosis: Based on the controlled attenuation parameter (CAP), it is defined as CAP ≥ 274 dB / m; Severe fatty degeneration: defined as CAP ≥ 302 dB / m; In addition, the risk of liver fibrosis can be assessed (based on liver stiffness measurement (LSM)) as follows: Significant fibrosis (≥F2): LSM ≥8.2kPa; Advanced fibrosis (≥F3): LSM ≥9.7kPa; Cirrhosis (F4): LSM ≥13.6 kPa; High-risk MASH (At-Risk MASH): defined as a FibroScan-AST (FAST) score >0.35, this indicator identifies individuals with active steatohepatitis and significant fibrosis; Specifically, the primary outcome of this protocol is MASLD, and secondary outcomes include non-alcoholic fatty liver disease (NAFLD), high-risk MASH, and severe MASLD with severe steatosis.

[0045] Step 3: Variable Selection On the training set, the aforementioned 17 candidate anthropometric / metabolic indicators were simultaneously incorporated into the LASSO logistic regression model, with the presence or absence of MASLD as the dependent variable. By adjusting the LASSO penalty parameter, the model automatically selected the combination of variables and their coefficients that best explain MASLD. The results identified several key predictors, including WC, WHtR, WHpR, TyG, CI, and LAP (preliminary results also include the derived indicator TyG). WHpR (which was later removed due to collinearity issues) The above-mentioned variables are statistically significantly correlated with MASLD and have low collinearity with each other, making them suitable as components of the composite index. The MBRI modeling process effectively eliminates redundant and collinear information, thus giving the model excellent robustness. On the independent validation set, the MBRI index shows a performance level comparable to that of the training set, which fully demonstrates that the model does not have an overfitting problem and has high reliability for general application.

[0046] Step 4: Model Building and MBRI Formula After obtaining the key variables, the above variables are combined according to their LASSO regression coefficients to define a new compliance assessment index, MBRI: MBRI=0.0073×WC+4.2025×WHtR+2.1155×WHpR+0.6214×TyG−0.0023×LAP+0.1504×CI; Step 5: MBRI Performance Verification After constructing the MBRI, its diagnostic performance was systematically evaluated on the training and validation sets. The receiver operating characteristic (ROC) curve and area under the curve (AUC) were mainly used to measure the MBRI's ability to distinguish between MASLD patients and non-patients. At the same time, calibration curves were plotted to test the consistency between the MBRI's predicted risk and the actual incidence rate, and decision curve analysis (DCA) was used to evaluate its clinical net benefit at different risk thresholds.

[0047] The results show that MBRI exhibits good discriminative power and calibration accuracy on both the training and validation sets. For example, the validation set results are as follows: Figure 3 As shown, the AUC of MBRI for MASLD reached 0.738, which was significantly higher than the AUC of the single indicator (0.660-0.712). Figures 4-6The graphs represent the operating characteristic (ROC) curves for subjects under other outcomes (NAFLD), at-risk MASH, and severe MASLD, respectively. The horizontal axis represents specificity, and the vertical axis represents sensitivity. Different colors in the legend correspond to different predictive indicators, and the corresponding AUC is given. Figures 4-6 As can be seen from the results, by comparing the discrimination ability of MBRI with the six component indicators in each outcome, it can be concluded that the AUC of the MBRI constructed in this embodiment for each outcome is greater than 0.7, and is higher than the AUC of other single indicators.

[0048] like Figures 7-10 As shown, Figures 7-10 The DCA curves for MASLD, NAFLD, at-risk MASH, and severe MASLD are shown below. The left side represents the test set, and the right side represents the training set. The horizontal axis represents the risk threshold, and the vertical axis represents the net benefit. In the legend, the black curve represents MBRI, and the other colored curves represent single indicators (WC, WHtR, WHpR, TyG, LAP, and CI, respectively). The red line "All" represents all intervention, and the blue line "None" represents no intervention. As can be seen from the graph, within most risk threshold ranges, MBRI has a higher or similar net benefit compared to single indicators. Multidimensional validation also demonstrates that the predicted probability of MBRI is in high agreement with the actual prevalence. The calibration curve is close to the 45° ideal line, and no obvious overfitting is observed. The DCA curves show that within the clinically relevant risk threshold range, MBRI has a higher net benefit compared to no intervention or using single indicators for decision-making. It can accurately quantify risk and guide reasonable intervention decisions, thus bringing real practical value in clinical scenarios and laying the foundation for its application as a diagnostic tool.

[0049] Step 6: Online Platforms and Tools To facilitate widespread application, this example also includes a supporting online MBRI calculation platform. Based on the Shiny architecture, it integrates user input with model calculation. Doctors or individuals only need to input raw data such as height, weight, waist circumference, blood glucose, and blood lipids into the platform to obtain MBRI index values ​​and corresponding MASLD risk assessment results in real time. The online tool makes MBRI calculation simple and fast, eliminating the need for users to manually input formulas or perform programming calculations. This invention can be conveniently used in both clinical and research settings. Furthermore, the index and calculation tool can be further integrated into hospital information systems or mobile health applications to achieve automated MASLD risk screening for individuals undergoing physical examinations. In summary, the process of this invention, from data modeling to tool development, forms a complete methodological chain, providing a practical solution for intelligent diagnosis of MASLD. It is convenient for clinical and personal use and significantly lowers the barrier to MASLD risk assessment.

[0050] Step 7: Machine Learning Model Building Using the MBRI joint laboratory index as the independent variable and MASLD as the dependent variable, variable selection (lasso combined with boruta) was performed. Finally, common blood indicators such as UA, HDL, and GHb, as well as MBRI, were included. Eight machine learning models were then established and compared with common NIT indicators. Cross-validation was used to evaluate the performance of the models, including AUC, sensitivity, specificity, PPV, NPV, and F1-score. The SHAP interpretation method was used to calculate feature importance and local / global contribution maps for each model, demonstrating the core role of MBRI in the model.

[0051] Furthermore, a MASLD risk prediction model was established using algorithms such as Gradient Boosting Machine (GBM), Logistic Regression (GLM), and Neural Network (NNET). Figures 11-14 As shown, Figures 11-14 The ROC curves for MASLD, NAFLD, at-risk MASH, and severe MASLD are shown respectively. The curves compare the discrimination ability of various machine learning models (GLM, RF, SVM, XGB, KNN, GBM, NNET, NB, etc.) built based on MBRI and biochemical indicators with traditional non-invasive scoring (FIB-4, NFS, HSI, APRI). The horizontal axis is 1-specificity, and the vertical axis is sensitivity. The AUC of each model is also shown in the figure. Figures 15-17 The SHAP beeswarm plots for the GLM, GBM, and NNET models are given, where the horizontal axis represents the SHAP value and the vertical axis represents each predictor variable (MBRI_8, HDL, UA, GHb). The color of the points from blue to red indicates that the variable values ​​are increasing. Figures 18-20 The bar chart of the mean absolute SHAP values ​​for the corresponding models shows the ranking of feature importance. As can be seen from the chart, MBRI_8 is the variable with the highest contribution in all three models, followed by GHb, HDL, and UA. Feature selection results show that MBRI remains one of the most important features (MBRI contributes the most in the SHAP interpretation analysis). Furthermore, the AUC of the model is further improved after incorporating blood indicators (it increased to over 0.75 on the validation set, a significant gain compared to using MBRI alone). This ensemble approach highlights the value of MBRI as a core feature, further improving the accuracy of MASLD hierarchical prediction.

[0052] Compared to existing single-indicator methods, MBRI can more accurately identify MASLD patients. Based on validation set data, the AUC of the MBRI index reached 0.738. Through DeLong's test, MBRI showed statistically significantly higher discriminative power than each of its individual components. In diagnosing NAFLD, at-risk MASLD, and severe MAFLD, the AUC of the MBRI index reached 0.736, 0.768, and 0.782, respectively. This means that using MBRI for screening can identify more true patients and reduce false positives. This superior diagnostic efficacy will directly translate into clinical benefits, allowing more high-risk individuals for MASLD to be identified in a timely and accurate manner, thus enabling them to receive further examination and intervention earlier. For general population screening in areas with high disease prevalence, MBRI is expected to serve as a reliable initial screening tool to improve early detection rates.

[0053] Through the above technical solutions, the MBRI index can also be integrated as a module into more complex prediction systems. The machine learning model constructed by combining MBRI with blood biochemical markers further improves the prediction accuracy of MASLD-related diseases. MBRI effectively extracts vital signs information closely related to MASLD, and has strong explanatory power and transfer value. MBRI can also be extended as needed to predict the progression risk of MASLD (such as developing into hepatitis MASH or the occurrence of fibrosis), and can be combined with other risk scoring systems to form a more comprehensive liver health assessment scheme. This invention leaves room for further algorithm improvement and functional expansion. Its core ideas and methods can continue to evolve in related fields, thereby maintaining the technological advancement and vitality.

[0054] Understandably, the information required for MBRI index calculation comes entirely from routine physical measurements and basic laboratory tests, without the need for specialized imaging or expensive testing methods. This makes MBRI highly suitable for use in primary healthcare and large-scale population screening. Once the model formula is determined, the calculation process involves only simple algebraic operations. Combined with the developed online calculation tools, users can master it with almost no additional training. Compared to liver ultrasound, CT scans, and other examinations, MBRI achieves zero-invasive, low-cost initial screening for MASLD. Compared to complex gene testing or metabolomics methods, MBRI has the advantages of convenient access and real-time results. Therefore, this invention has good accessibility and cost-effectiveness, facilitating its promotion and application in medical institutions at all levels. Through internet platforms and mobile devices, MBRI can also be embedded in personal health management applications to help high-risk groups self-monitor their liver health. This ease of use and low cost will greatly promote the expansion of MASLD screening from professional medical fields to the field of public health management, resulting in significant social benefits.

[0055] It should be noted that this embodiment provides a risk assessment method for metabolic-related fatty liver disease (MASLD). By using LASSO regression for variable screening, a MASLD risk prediction model is constructed. Through regularization screening of key features, redundant or highly collinear variables are eliminated, ensuring the simplicity and stability of the model. This effectively avoids the interference of multicollinearity on the diagnostic index, improves the model's generalization performance and interpretability. By setting the diagnostic threshold of MBRI or directly using the incidence probability corresponding to MBRI, doctors can more accurately distinguish high-risk patients from the general population. This discrimination rule significantly improves the sensitivity and specificity of MASLD screening, more reliably identifies high-risk patients of MASLD, and reduces the rate of missed diagnoses and misdiagnoses.

[0056] Example 2 Please see Figure 2 This embodiment provides a risk assessment system for metabolic-related fatty liver disease, the system comprising: Acquisition module 100: used to acquire multiple initial anthropometric parameters and initial metabolic parameters of the target object, and to determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; Indicator filtering module 200: used to filter out a target indicator combination from the plurality of candidate indicators; Risk assessment module 300: used to determine an assessment index for assessing the risk of metabolic-associated fatty liver disease based on the target indicator combination, and to conduct a risk assessment of metabolic-associated fatty liver disease on the target subject according to the assessment index.

[0057] Furthermore, in this embodiment, the indicator screening module is specifically used to perform feature selection on the multiple candidate indicators using the LASSO regression algorithm, so as to screen out the target indicator combination including waist circumference, waist-to-height ratio, waist-to-hip ratio, triglyceride-glucose index, conic index and lipid accumulation products, and determine the weight coefficient corresponding to each indicator. Then, the indicators in the target indicator combination and their corresponding weight coefficients are linearly combined to generate the target index.

[0058] It should be noted that this embodiment provides a risk assessment system for metabolic-associated fatty liver disease (MASLD), including an acquisition module 100, an indicator screening module 200, and a risk assessment module 300. Among them, the assessment index MBRI, with its higher accuracy, good robustness, and convenient accessibility, effectively solves the technical problems of low diagnostic efficiency and lack of unified standards in the prior art for MASLD, significantly improves the non-invasive risk assessment level of MASLD, and is expected to become an important tool in clinical practice and public health screening, with significant clinical value and industrial prospects.

[0059] In a preferred embodiment, this application also provides an electronic device, the electronic device comprising: The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the risk assessment method for metabolic-related fatty liver disease. The computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the method of the present invention.

[0060] This invention can also be implemented as an online MBRI calculation platform. Based on the Shiny architecture, it integrates user input and model calculation. Doctors or individuals only need to input raw data such as height, weight, waist circumference, blood glucose, and blood lipids into the platform to obtain MBRI index values ​​and corresponding MASLD risk assessment results in real time. The provision of online tools makes MBRI calculation simple and fast. Users do not need to manually substitute formulas or perform programming calculations. The results of this invention can be conveniently used in clinical practice and scientific research.

[0061] This invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0062] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0063] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0064] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A risk assessment method for metabolic-associated fatty liver disease, characterized in that, include: S100: Obtain multiple initial anthropometric parameters and initial metabolic parameters of the target object, and determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; S200. Select a target indicator combination from the plurality of candidate indicators; S300. Based on the target indicator combination, determine the assessment index for evaluating the risk of metabolic-related fatty liver disease, and conduct a risk assessment of metabolic-related fatty liver disease on the target subject according to the assessment index.

2. The risk assessment method for metabolic-related fatty liver disease according to claim 1, characterized in that, The method for selecting a combination of target indicators from the plurality of candidate indicators includes: The LASSO regression algorithm is used to perform feature selection on the multiple candidate indicators in order to screen out the target indicator combination and determine the weight coefficients corresponding to each target indicator.

3. The risk assessment method for metabolic-related fatty liver disease according to claim 2, characterized in that, The method for selecting the combination of target indicators and determining the weight coefficients corresponding to each target indicator includes: Using the multiple candidate indicators as independent variables and the prevalence of metabolic-related fatty liver disease as the dependent variable, a LASSO regression model was constructed. By adjusting the penalty parameter of the LASSO regression model, a subset of target indicators is selected from the multiple candidate indicators; Based on the regression coefficients of each indicator in the LASSO regression model within the target indicator subset, the weight coefficients corresponding to each target indicator are determined.

4. The risk assessment method for metabolic-related fatty liver disease according to claim 3, characterized in that, The process of selecting the target indicator combination also includes: Multicollinearity is judged on the target indicator combination obtained by the LASSO regression algorithm. Based on the judgment result, derivative indicators that are collinear with at least one other target indicator are removed from the target indicator combination obtained by the screening. The derived index is a composite index calculated from at least two initial anthropometric parameters or initial metabolic parameters.

5. The risk assessment method for metabolic-related fatty liver disease according to claim 4, characterized in that, The target indicator combination includes: Waist circumference, waist-to-height ratio, waist-to-hip ratio, triglyceride-glucose index, conic index, and lipid accumulation products.

6. The risk assessment method for metabolic-related fatty liver disease according to claim 5, characterized in that, The method for determining the assessment index for evaluating the risk of metabolic-related fatty liver disease based on the target indicator combination includes: The evaluation index is obtained by taking each index in the target index combination after multicollinearity processing as input variables, and taking the weight coefficients corresponding to each target index determined by the LASSO regression algorithm as weight coefficients of each input variable, and performing a weighted summation operation.

7. The risk assessment method for metabolic-related fatty liver disease according to claim 6, characterized in that, The obtained evaluation index is represented as follows: MBRI=k1×WC+k2×WHtR+k3×WHpR+k4×TyG−k5×LAP+k6×CI; Wherein, MBRI represents the assessment index, WC represents waist circumference, k1 represents the weighting coefficient of waist circumference, WHtR represents waist-to-height ratio, k2 represents the weighting coefficient of waist-to-height ratio, WHpR represents waist-to-hip ratio, k3 represents the weighting coefficient of waist-to-hip ratio, TyG represents the triglyceride-glucose index, k4 represents the weighting coefficient of triglyceride-glucose index, LAP represents lipid accumulation products, k5 represents the weighting coefficient of lipid accumulation products, CI represents the conic index, and k6 represents the weighting coefficient of conic index.

8. The risk assessment method for metabolic-related fatty liver disease according to claim 1, characterized in that, The risk assessment of metabolic-related fatty liver disease in the target subject based on the assessment index includes: The assessment index is compared with a preset threshold, and the risk assessment result of the target object regarding metabolic-related fatty liver disease is output based on the comparison result.

9. A risk assessment system for metabolic-associated fatty liver disease, characterized in that, include: Acquisition module: used to acquire multiple initial anthropometric parameters and initial metabolic parameters of the target object, and to determine multiple candidate indicators based on the multiple initial anthropometric parameters and initial metabolic parameters; Indicator filtering module: used to filter out a target indicator combination from the multiple candidate indicators; Risk assessment module: used to determine an assessment index for assessing the risk of metabolic-associated fatty liver disease based on the target indicator combination, and to conduct a risk assessment of metabolic-associated fatty liver disease on the target subject according to the assessment index.

10. An electronic device, characterized in that, include: Memory; The processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the risk assessment method for metabolic-associated fatty liver disease according to any one of claims 1 to 8.

Images