A method and system for multiple liver pathological evaluations based on machine learning

Abstract

The present invention provides a method and system for evaluating various liver pathologies based on machine learning, which obtains the weight information, clinical examination information, and evaluation results of liver tissue samples of multiple historical patients; based on the obtained information, performs data preprocessing to construct a training set and a test set; uses the training set to train Logistic regression and Ridge regression models to determine the performance of the models in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH, and determines the final prediction index parameters and prediction models; obtains the weight information, clinical examination information, and evaluation results of liver tissue samples of a patient, extracts the numerical values related to the prediction index parameters therefrom, and processes them with the determined trained prediction models for the corresponding prediction index parameters to obtain the final prediction results. The present invention constructs a simple model by obtaining clinical indicators and can comprehensively evaluate all aspects of the liver pathology of chronic liver diseases.

Description

Technical Field

[0001] The present invention belongs to the field of pathological evaluation, and particularly relates to a method and system for evaluating multiple liver pathologies based on machine learning. Background Art

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

[0003] Studies have shown that chronic liver diseases are closely related to the occurrence and development of various extrahepatic diseases, including but not limited to obesity, diabetes, and cardiovascular diseases (CVD). Large-scale cohort studies have shown that whether it is simple fatty liver, non-fibrotic metabolic dysfunctional steatohepatitis (MASH), or fibrotic state that has not reached the stage of cirrhosis, as metabolic associated fatty liver disease (MAFLD) progresses, the risk of all-cause mortality in patients increases significantly. Therefore, for chronic liver diseases, early diagnosis and intervention are of great significance for improving prognosis and reducing complications.

[0004] Currently, liver biopsy remains the gold standard for diagnosing chronic liver diseases and evaluating histological changes. Early detection of histological changes in the liver can create opportunities to delay the progression of the disease and bring significant social and economic benefits. Although many current studies focus on blood-based omics detection as the main approach for non-invasive detection, with the expectation of achieving early diagnosis of chronic liver diseases, there are still limitations in terms of accessibility and clinical translation. In addition, liver pathological diagnosis includes multiple indicators such as steatosis, lobular inflammation, ballooning degeneration of hepatocytes, and fibrosis. For each liver pathological indicator, different models may need to be constructed to ensure the accuracy of prediction, which also increases the difficulty of application. Therefore, there is an urgent need for a set of easy-to-use, low-cost, sensitive, accurate, and comprehensive markers that can uniformly predict these characteristics and then efficiently predict the occurrence of chronic liver diseases.

[0005] Machine learning has become a powerful tool for predicting and managing complex diseases. Through training on large-scale clinical and multi-omics data, machine learning models can identify complex biomarkers that are crucial for early diagnosis and risk stratification. However, these models need to learn from comprehensive datasets to identify the relationships between different parameters and outcomes, so as to accurately predict the occurrence and development of chronic liver diseases. Therefore, although machine learning has great potential in deepening our understanding and prediction of such diseases, due to the lack of large-scale prospective studies integrating multiple types of clinical data, the clinical translation of machine learning tools in the prediction and diagnosis of MAFLD is hindered. Summary of the Invention

[0006] To solve the above problems, the present invention proposes a method and system for evaluating various liver pathologies based on machine learning. By obtaining clinical indicators to construct a simple model, the present invention can comprehensively evaluate all aspects of the liver pathology of chronic liver diseases.

[0007] According to some embodiments, the present invention adopts the following technical solutions:

[0008] A method for evaluating various liver pathologies based on machine learning, comprising the following steps:

[0009] Obtain the weight information, clinical examination information, and pathological evaluation results of liver tissue samples of multiple historical patients;

[0010] Based on the obtained information, perform data preprocessing, construct a training set and a test set, and determine the key prediction features of each pathological index;

[0011] Use the training set to train a machine learning model, determine the performance of the model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH, and determine the final prediction index parameters and prediction model;

[0012] Obtain the weight information, clinical examination information, and evaluation results of liver tissue samples of a patient, extract the relevant numerical values of the prediction index parameters, and process them with the determined trained prediction model for the corresponding prediction index parameters to obtain the final prediction result.

[0013] As an alternative embodiment, the weight information includes BMI data; the anthropometric parameters include gender, age, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), waist circumference, hip circumference, and waist-to-hip ratio (WHR); the complete blood count includes white blood cell count (WBC), neutrophil ratio (Neu ratio), lymphocyte ratio (Lym ratio), eosinophil ratio (EOS ratio), basophil ratio (BAS ratio), monocyte ratio (Mono ratio), neutrophil count (Neu), lymphocyte (Lym), eosinophil (EOS), basophil (BAS), monocyte (Mono), red blood cell count (RBC), hemoglobin (Hb), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), red blood cell distribution width (RDW), platelet count (PLT), platelet distribution width (PDW), mean platelet volume (MPV), and plateletcrit (PCT); the biochemical indices include prealbumin (PA), total protein (TP), albumin (ALB), globulin (GLB), albumin / globulin ratio (AGR), alpha-1 antitrypsin (Alpha.1), alpha-2 macroglobulin (Alpha.2), beta-1 globulin (Beta.1), beta-2 microglobulin (Beta.2), gamma globulin (Gamma), alanine aminotransferase (ALT), aspartate aminotransferase (AST), glutamate dehydrogenase (GDH), gamma-glutamyl transferase (GGT), alkaline phosphatase (AKP), adenosine deaminase (ADA), alpha-L-fucosidase (AFU), total bilirubin (TBIL), direct bilirubin (DBIL), indirect bilirubin (IBIL), total bile acid (TBA), homocysteine (Hcy), phospholipase A2 (PLA2), beta-hydroxybutyric acid (HBUT), urea, creatinine (Cr), uric acid (UA), complement component 1q (C1q), creatine kinase (CK), creatine kinase MB isoenzyme (CK.MB), cardiac troponin I (CTNI), lactate dehydrogenase (LDH), folic acid (FOL), and angiotensin II (AII); the blood lipid indices include total cholesterol (TC), high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), small dense low-density lipoprotein cholesterol (sdLDL), apolipoprotein A1 (ApoA1), apolipoprotein B (ApoB), triglyceride (TG), lipoprotein A (LP.a), non-esterified free fatty acid (NEFA); the blood glucose indices include fasting blood glucose (GLU), glycated albumin index (GA.index), ischemia modified albumin (IMA), sialic acid (SA), glycated hemoglobin (HbA1c), fasting insulin (FINS), fasting C-peptide (FCP), homeostasis model assessment of insulin resistance (HOMA-IR), adipose insulin resistance (Adipo-IR), triglyceride glucose index (TyG), triglyceride glucose body mass index (TyG-BMI), and triglyceride glucose waist circumference index (TyG-WC); the coagulation indexes include prothrombin time (PT-S), international normalized ratio of prothrombin time (PT-INR), prothrombin activity (PT%), prothrombin time ratio (PT-R), activated partial thromboplastin time (APTT-S), activated partial thromboplastin time ratio (APTT-R), fibrinogen (FIB), thrombin time (TT-S), thrombin time ratio (TT-R), and D-dimer (DD.i); the metal indexes include calcium (Ca), cadmium (Cd), cobalt (Co), chromium (Cr), copper (Cu), iron (Fe), lithium (Li), magnesium (Mg), manganese (Mn), nickel (Ni), lead (Pb), and zinc (Zn); the endocrine hormone indexes include prolactin (PRL), testosterone (T), luteinizing hormone to follicle-stimulating hormone ratio (LH / FSH), growth hormone (hGH), free triiodothyronine (FT3), free thyroxine (FT4), and thyroid-stimulating hormone (TSH).

[0014] As an alternative embodiment, the pathological evaluation results of the liver tissue sample include ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH.

[0015] As an alternative embodiment, the process of data preprocessing includes:

[0016] In the training set, the original data is subjected to z-score transformation according to the mean and standard deviation of each index to achieve data standardization. The formula for z-score transformation is:

[0017] z = (x - μ) / σ;

[0018] where z is the z-score value of each data, that is, the value after standardization; x is the original value of the data; μ is the mean of each index in the original data; σ is the standard deviation of each index in the original data.

[0019] As an alternative embodiment, during the preprocessing of the test set data, each clinical index is standardized using the mean and standard deviation of the training set, that is, subtracting the mean of the training set and dividing by the standard deviation of the training set as the standardized value to achieve the standardization of the test data.

[0020] As an alternative implementation, during the data preprocessing, samples with various pathological outcomes were labeled as 1, and samples without an outcome were labeled as 0 for input into the machine learning model.

[0021] As an alternative implementation, in the training set, the best predictive features for each liver pathological outcome were screened through recursive feature elimination.

[0022] As an alternative implementation, 23 common machine learning algorithms were used to build models to determine the performance of each model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk of NASH. Based on balancing model accuracy and clinical utility, we screened out the optimal algorithm and the number of features for each liver pathological result, thereby determining the final predictive index parameters and prediction models for each pathological outcome. Meanwhile, the performance of the models was tested in internal and external test cohorts;

[0023] As an alternative implementation, for the Logistic regression model, a regularization term was added to prevent overfitting, and the loss function of the model was set as:

[0024] J(w) = -(1 / N)Σ[y i *log(p i ) + (1 - y i )*log(1 - p i )] + λ*R(w);

[0025] where J(w) is the loss function; N is the total number of samples; y i is the actual label (0 or 1) of the i-th sample; p i is the probability that the i-th sample is predicted as class 1; λ is the regularization strength, automatically selected by cross-validation; R(w) is the regularization term, using L1 regularization;

[0026] The prediction formula of the model is:

[0027] p(y = 1|x) = σ(w0 + Σ(w j *x j ))

[0028] where: w0 is the bias term, output by the model; w j is the weight, that is, the contribution of each feature to the model, output by the model; x j is the j-th component of the input feature vector, determined by the specific numerical values of the indicators of each sample in the training set;

[0029] If p(y = 1|x) ≥ 0.5, predict that the sample is of class 1;

[0030] If p(y = 1|x) < 0.5, predict that the class of this sample is 0.

[0031] As an alternative implementation, for the Ridge regression model, define its loss function as:

[0032] L(w) = Σ(y i -(w0 + Σ(w j *x ij ))) 2 + α * Σ(w j 2 )

[0033] where y i is the actual label of the i-th sample; w0 is the bias term of the model; w j is the weight parameter of the j-th feature; x ij is the j-th feature value of the i-th sample; α is the regularization strength;

[0034] The prediction formula of the model is:

[0035]

[0036] where w j is the weight coefficient of the j-th feature, output by the model; xi j is the j-th feature value of the i-th sample, determined by the specific numerical values of the indicators of each sample in the training set.

[0037] If predict that the class of this sample is 1;

[0038] If predict that the class of this sample is 0.

[0039] Successively establish machine learning models for ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH, and output the predicted class of each sample by the model, that is, 0 or 1;

[0040] Under repeated stratified 5-fold cross-validation, calculate the balanced accuracy, precision, recall, F1 score, and AUROC of each model;

[0041] The calculation methods of each index are as follows:

[0042] Balanced accuracy = (1 / 2) * [(TP / (TP + FN)) + (TN / (TN + FP))];

[0043] Precision = TP / (TP + FP);

[0044] Recall = TP / (TP + FN);

[0045] F1 score = 2 * (Precision * Recall) / (Precision + Recall);

[0046] AUROC = ∫(curve of TPR against FPR);

[0047] Wherein, TP is the true positive, TN is the true negative, FN is the false negative, and FP is the false positive.

[0048] As an alternative implementation, the final selected prediction metrics include, for example Figure 3 as shown:

[0049] For ballooning degeneration, alanine aminotransferase (ALT), total bile acid (TBA), and homeostasis model assessment of insulin resistance (HOMA-IR) are selected as prediction metrics, and the Ridge regression model is selected;

[0050] For steatosis, ALT and glutamate dehydrogenase (GDH) are selected as prediction metrics, and the Ridge regression model is selected;

[0051] For lobular inflammation, ALT, prothrombin activity (PT%), and cadmium (Cd) are selected as prediction metrics, and the Logistic regression model is selected;

[0052] For significant fibrosis, peripheral blood monocyte ratio (Mono.ratio), aspartate aminotransferase (AST), FPG, glycated albumin index (GA.index), and β2-microglobulin (Beta2) are selected as prediction metrics, and the Logistic regression model is selected;

[0053] For NASH, phospholipase A2 (PLA2), Cd, magnesium (Mg), fasting insulin (FINS), HOMA-IR, and adipose tissue insulin resistance index (Adipo-IR) are selected as prediction metrics, and the Logistic regression model is selected;

[0054] For NAS score, ALT, cobalt (Co), and triglyceride glucose body mass index (TyG-BMI) are selected as prediction metrics, and the Logistic regression model is selected;

[0055] For risk NASH, ALT, GA.index, and Beta2 are selected as prediction metrics, and the Logistic regression model is selected.

[0056] Figure 3 The left side in [[]] shows the ROC-AUC of each model in the training group and the test group, and the right side shows the recall rate of each model. The right table lists the features used by each model.

[0057] Furthermore, from the seven independent models for each pathological outcome, ALT, HOMA-IR, and Mono.ratio were selected as key clinical indicators, and a Logistic regression model was constructed to simultaneously predict multiple liver pathological outcomes as the prediction target.

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] By obtaining simple and easily accessible clinical indicators to construct a machine learning model, the present invention can comprehensively evaluate all aspects of the liver pathology of chronic liver diseases and improve the efficiency and performance of pathological prediction.

[0060] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given in conjunction with the accompanying drawings and are described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0062] Figure 1 It is a schematic flow chart of the method for an embodiment;

[0063] Figure 2 It is a schematic flow chart of the method for another embodiment.

[0064] Figure 3 It is the performance of different prediction models for various liver pathological types.

[0065] Figure 4 It is a simplified general model for predicting multiple liver pathological outcomes.

[0066] Figure 5 It is the performance of the simplified model in the external validation cohort in China.

[0067] Figure 6 It is the performance of the simplified model in the NHANES cohort in the United States. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0069] It should be noted that the following detailed descriptions are all illustrative and are intended to provide a further description of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0070] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0071] In the case of no conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0072] Embodiment 1

[0073] A method for evaluating multiple liver pathologies based on machine learning, as Figure 1 shown, includes the following steps:

[0074] A data acquisition module that acquires the weight information, clinical examination information, and evaluation results of liver tissue samples of multiple historical patients;

[0075] A preprocessing module that performs data preprocessing based on the acquired information to construct a training set and a test set;

[0076] A model training and validation module that trains 23 different machine learning models using the training set to determine the performance of each model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH. On the basis of balancing model accuracy and clinical practicability, we screened out the optimal algorithm and the number of features for each liver pathology result, thereby determining the final prediction index parameters and prediction models for each pathological outcome. At the same time, the performance of the model was tested in internal and external test cohorts;

[0077] A prediction module that acquires the weight information, clinical examination information, and evaluation results of liver tissue samples of a patient, extracts the numerical values related to the prediction index parameters therein, and processes them with the determined trained prediction model for the corresponding prediction index parameters to obtain the final prediction result.

[0078] The following is an introduction step by step.

[0079] The data acquisition module aims to acquire the weight information, clinical examination information, and evaluation results of liver tissue samples of multiple historical patients.

[0080] In this embodiment, the training cohort includes 909 individuals who underwent sleeve gastrectomy for the treatment of morbid obesity at Qilu Hospital of Shandong University from May 2018 to October 2023.

[0081] The test cohort consisted of 186 individuals who underwent sleeve gastrectomy at the same hospital between December 2023 and June 2024. Liver tissue samples from the patients were collected during the operation to evaluate liver injury and pathological changes in the liver.

[0082] Physical examinations, clinical, and laboratory examinations of the human body were all conducted 1 - 2 days before the operation. The body mass index (BMI) is calculated as weight (kg) / height (m). 2 . The homeostasis model assessment of insulin resistance (HOMA-IR) is calculated as fasting glucose (mmol / L) × fasting insulin (μIU / mL) / 22.5.

[0083] The diagnostic criteria for diabetes are a fasting plasma glucose (FPG) level ≥7.0 mmol / L or a glycated hemoglobin (HbA1c) level ≥6.5% (48 mmol / mol). The diagnosis of metabolic syndrome is based on the CDS2004 criteria, and an individual needs to meet at least three of the following diagnostic criteria: body mass index (BMI) ≥25 kg / m 2 ; blood pressure ≥140 / 90 mmHg or being on antihypertensive medications; fasting plasma glucose ≥6.1 mmol / L and / or 2-hour plasma glucose during an oral glucose tolerance test (OGTT) ≥7.8 mmol / L, or being on hypoglycemic medications; triglycerides (TG) ≥1.7 mmol / L and / or high-density lipoprotein cholesterol (HDL-c) <0.9 mmol / L in men and <1.0 mmol / L in women.

[0084] Liver biopsy samples were systematically evaluated by experienced liver pathologists, and the analysis of the training set and the test set was performed by different experts separately. Liver fibrosis was graded using a 5-point scale (0 to 4 points). According to the NASH CRN guidelines and the NAFLD activity score system (NAS), histological changes - steatosis, lobular inflammation (0 to 3 points), and ballooning (0 to 2 points) - were evaluated. The NAS score is the sum of the scores for steatosis, lobular inflammation, and ballooning, ranging from 0 to 8 points.

[0085] The preprocessing module, and the process of data preprocessing includes:

[0086] In the training set, the raw data was z-score transformed based on the mean and standard deviation of each metric to achieve data standardization. The formula for z-score transformation is:

[0087] z = (x - μ) / σ;

[0088] where z is the z-score value of each data, that is, the standardized value; x is the raw value of the data; μ is the mean of each metric in the raw data; σ is the standard deviation of each metric in the raw data.

[0089] During the data preprocessing process, for the test cohort, each clinical indicator is standardized using the mean and standard deviation of the training set, that is, subtracting the mean of the training set and dividing by the standard deviation of the training set as the standardized value to achieve the standardization of the test data.

[0090] During the process of data preprocessing, the samples with each pathological outcome are marked as 1, and the samples without an outcome are marked as 0 for input into the machine learning model.

[0091] The baseline information of the training set and the test set is as follows: the average age of the cohort is 31.5 years old, and the average body mass index (BMI) is 41.85 kg / m 2 . Among the study subjects, 29.6% are male, 35.7% have type 2 diabetes (T2D), and 48% have metabolic syndrome (MetS). Liver pathology data shows that 89.8% of people show ballooning degeneration (ballooning degeneration score > 0), 73.9% have steatosis (steatosis score > 0), 60.9% have lobular inflammation (inflammation score > 0), and 22.2% have significant fibrosis (fibrosis score ≥ 2). A total of 8.4% are diagnosed with NASH, 27.5% have a NAS score ≥ 5, and 14.1% are NASH with fibrosis, called "risk NASH" (NASH with a fibrosis score ≥ 2).

[0092]

[0093] Model training and validation module

[0094] In this embodiment, when performing feature screening, in the training set, we use recursive feature elimination to screen the best predictive features for each liver pathological outcome. Specifically, Logistic regression is used as the basic model for fitting the data, and 1 feature is removed in each recursion to obtain the optimal predictive features for each liver pathological outcome.

[0095] In this embodiment, we use 23 different algorithms to train the machine learning model and adopt the method of repeated stratified 5-fold cross-validation to evaluate the performance of the model by calculating the balanced accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve (AUROC). Based on balancing the model accuracy and clinical practicability, we screen out the optimal algorithm and the number of features for each liver pathological result.

[0096] For all seven liver pathological outcomes, 23 models of ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH are established in sequence, and the predicted class of each sample by the model, that is, 0 or 1, is output;

[0097] Under repeated stratified 5-fold cross-validation, the balanced accuracy, precision, recall, F1-score, and AUROC of each model were calculated;

[0098] The calculation methods for each metric are as follows:

[0099] Balanced accuracy = (1 / 2) * [(TP / (TP + FN)) + (TN / (TN + FP))];

[0100] Precision = TP / (TP + FP);

[0101] Recall = TP / (TP + FN);

[0102] F1-score = 2 * (Precision * Recall) / (Precision + Recall);

[0103] AUROC = ∫(curve of TPR against FPR);

[0104] Where TP is the true positive, TN is the true negative, FN is the false negative, and FP is the false positive.

[0105] Based on balancing model accuracy and clinical utility, the optimal number of features and the optimal model were selected for each liver pathological result.

[0106] For ballooning degeneration, we selected alanine aminotransferase (ALT), total bile acid (TBA), and homeostasis model assessment of insulin resistance (HOMA-IR), and established a model through Ridge regression, with an AUROC of 0.72 (95% CI: 0.67 - 0.76) and a recall of 0.52 (95% CI: 0.48 - 0.56) on the training set;

[0107] For steatosis, we selected ALT and glutamate dehydrogenase (GDH), and established a model through Ridge regression, with an AUROC of 0.74 (95% CI: 0.72 - 0.77) and a recall of 0.55 (95% CI: 0.52 - 0.59) on the training set;

[0108] For lobular inflammation, we selected ALT, prothrombin activity (PT%), and cadmium (Cd), and established a model through Logistic regression, with an AUROC of 0.63 (95% CI: 0.58 - 0.68) and a recall of 0.59 (95% CI: 0.54 - 0.65) on the training set;

[0109] For significant fibrosis, we selected the proportion of peripheral blood mononuclear cells (Mono.ratio), aspartate aminotransferase (AST), FPG, glycated albumin index (GA.index), β2-microglobulin (Beta2), and established a model through Logistic regression. Its AUROC on the training set was 0.63 (95% CI: 0.58 - 0.67), and the recall rate was 0.48 (95% CI: 0.43 - 0.53);

[0110] For NASH, we selected phospholipase A2 (PLA2), Cd, magnesium (Mg), fasting insulin (FINS), HOMA-IR, adipose tissue insulin resistance index (Adipo-IR), and established a model through Logistic regression. Its AUROC on the training set was 0.66 (95% CI: 0.59 - 0.63), and the recall rate was 0.51 (95% CI: 0.44 - 0.59);

[0111] For NAS score, we selected alanine aminotransferase (ALT), cobalt (Co), triglyceride glucose body mass index (TyG-BMI), and established a model through Logistic regression. Its AUROC on the training set was 0.73 (95% CI: 0.70 - 0.76), and the recall rate was 0.62 (95% CI: 0.55 - 0.70);

[0112] For risk NASH, we selected ALT, GA.index, Beta2, and established a model through Logistic regression. Its AUROC on the training set was 0.69 (95% CI: 0.66 - 0.72), and the recall rate was 0.53 (95% CI: 0.46 - 0.60).

[0113] The Logistic regression and Ridge regression models involved were constructed in the following manner:

[0114] For the Logistic regression model, a regularization term was added to prevent overfitting, and the loss function of this model was set as:

[0115] J(w) = -(1 / N)Σ[y i *log(p i )+(1 - y i )*log(1 - p i )]+λ*R(w);

[0116] Among them, J(w) is the loss function; N is the total number of samples; y i is the actual label (0 or 1) of the i-th sample; p iThe probability that the i-th sample is predicted as class 1; λ is the regularization strength, automatically selected by cross-validation; R(w) is the regularization term, using L1 regularization;

[0117] The prediction formula of the model is:

[0118] p(y = 1|x) = σ(w0 + Σ(w j *x j ))

[0119] where: w0 is the bias term, output by the model; w j is the weight, that is, the contribution of each feature to the model, output by the model; x j is the j-th component of the input feature vector, determined by the specific numerical values of the indicators of each sample in the training set;

[0120] If p(y = 1|x) ≥ 0.5, predict that the sample is of class 1;

[0121] If p(y = 1|x) < 0.5, predict that the sample is of class 0.

[0122] For the Ridge regression model, define its loss function as:

[0123] L(w) = Σ(y i -(w0 + Σ(w j *x ij ))) 2 + α * Σ(w j 2 )

[0124] where, y i is the actual label of the i-th sample; w0 is the bias term of the model; w j is the weight parameter of the j-th feature; x ij is the j-th feature value of the i-th sample; α is the regularization strength;

[0125] The prediction formula of the model is:

[0126]

[0127] where, w j is the weight coefficient of the j-th feature, output by the model; xi j is the j-th feature value of the i-th sample, determined by the specific numerical values of the indicators of each sample in the training set.

[0128] If predict that the sample is of class 1;

[0129] If predict that the sample is of class 0.

[0130] In the test set, each clinical indicator was first standardized using the mean and standard deviation of the training set. Subsequently, we calculated the recall rate and AUROC of the model to evaluate its performance on the test set. Specifically, the AUROCs of the model in predicting ballooning degeneration, predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH were 0.66, 0.78, 0.56, 0.73, 0.64, 0.73, and 0.76 respectively

[0131] Comparison with existing models

[0132] The proposed model was compared with existing clinical models, including AAR, APRI, FCI, FI, FIB4, Fibroindex, Forns, GUCI, HGM1, HGM2, MODEL3, AP index, and CDS score (for predicting significant fibrosis), as well as LFS, HSI, and FLI (for predicting steatosis). For each participant, we first calculated the scores of each person using these models and trained them through a Logistic regression model. We evaluated the model performance by comparing the AUROCs in the test cohort to assess whether our model was superior to existing models. The AUROC results showed that in the test cohort, our model was consistently superior to existing prediction tools. For example, in predicting steatosis and significant fibrosis, the maximum AUROCs of existing models were 0.61 and 0.66 respectively, while the AUROCs of our machine learning model reached 0.71 in both predictions

[0133] Example 2

[0134] A machine learning-based multiple liver pathology assessment system, as Figure 2 shown, includes:

[0135] A data acquisition module for obtaining the detection values of ALT, HOMA-IR, and Mono.ratio of multiple historical patients, as well as the evaluation results of liver tissue samples

[0136] A preprocessing module for performing data preprocessing based on the obtained information to construct a training set and a test set

[0137] A model training and validation module for using the training set data and three clinical indicators of ALT, HOMA-IR, and Mono.ratio to establish a Logistic regression model. Determine the performance of the model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH; and test the model performance in internal and external test sets

[0138] A prediction module, which is used to obtain the specific values of three clinical indicators, namely ALT, HOMA-IR, and Mono.ratio, input them into the trained model, and obtain the final prediction result.

[0139] Specifically, to improve the clinical utility of the model, we manually selected three key clinical indicators, namely ALT, HOMA-IR, and Mono.ratio, from the seven independent models originally for each pathological outcome in Example 1, and tried to construct a simplified model that can predict multiple pathological outcomes simultaneously.

[0140] As Figure 4 shown, a simplified general model for predicting multiple liver pathological outcomes, in which A. Machine learning workflow for training and testing a logistic regression model for predicting multiple liver pathological outcomes. Training ROC, testing ROC, and confusion matrices for flatulence (B), steatosis (C), lobular inflammation (D), significant fibrosis (E), and NASH (F). TPR: True positive rate; FPR: False positive rate.

[0141] The training set and internal test set in Example 2 are the same as the data used in Example 1.

[0142] In addition, an external Chinese test cohort was established, and the data were from Qianfoshan Hospital in Shandong Province and Jiangsu Provincial People's Hospital. These two cohorts were also composed of obese individuals who underwent sleeve gastrectomy due to morbid obesity. Each patient signed an informed consent form.

[0143] The data of the NHANES cohort were from the examination data from 2017 to March 2020. Participants with missing relevant data were excluded from the analysis. The diagnosis of liver fibrosis and steatosis was performed by transient elastography of the liver ultrasound, where steatosis was defined as the controlled attenuation parameter (CAP) ≥ 274, and the criterion for significant fibrosis was the liver stiffness measurement (LSM) ≥ 8 kPa.

[0144] To avoid the problem of collinearity, among the pathological outcomes related to NASH, we selected NASH as the prediction target and excluded the risk of NASH and NAS scores. Subsequently, we constructed a Logistic regression model for the other five pathological outcomes based on these three indicators.

[0145] For the Logistic regression model, a regularization term was added to prevent overfitting, and the loss function of the model was set as:

[0146] J(w) = -(1 / N)Σ[y i *log(p i )+(1 - y i )*log(1 - p i )]+λ*R(w);

[0147] Among them, J(w) is the loss function; N is the total number of samples; y i is the actual label (0 or 1) of the i-th sample; p i is the probability that the i-th sample is predicted as class 1; λ is the regularization strength, automatically selected by cross-validation; R(w) is the regularization term, using L1 regularization;

[0148] The prediction formula of the model is:

[0149] p(y = 1|x) = σ(w0 + Σ(w j *x j ))

[0150] Among them: w0 is the bias term, output by the model; w j is the weight, that is, the contribution of each feature to the model, output by the model; x j is the j-th component of the input feature vector, determined by the specific numerical values of the indicators of each sample in the training set;

[0151] If p(y = 1|x) ≥ 0.5, predict that the sample is class 1;

[0152] If p(y = 1|x) < 0.5, predict that the sample is class 0.

[0153] In the training set, the AUROCs of the simplified model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH are 0.71, 0.75, 0.56, 0.60, and 0.63 respectively.

[0154] In the internal and external test sets, first standardize each clinical indicator using the mean and standard deviation of the training set.

[0155] Subsequently, calculate the recall rate and AUROC of the model to evaluate the performance of the model on the test set. In the internal test set, the AUROCs of the simplified model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH are 0.70, 0.66, 0.62, 0.65, and 0.71 respectively.

[0156] In the external test set, the model also shows performance similar to that of the training set and the internal test set. Such as Figure 5As shown, the ROC curves and confusion matrices of the Qianfoshan cohort in China (sample size = 217) show the diagnostic performance of the simplified model in identifying steatosis (A), lobular inflammation (B), significant fibrosis (C), and NASH (D). The ROC-AUCs and confusion matrices of the Jiangsu cohort (sample size = 103) show the performance of the model in identifying ballooning degeneration (E), steatosis (F), lobular inflammation (G), significant fibrosis (H), and NASH (I).

[0157] For the test cohort of Qianfoshan Hospital in Shandong Province, the AUROCs of the simplified model in predicting steatosis, lobular inflammation, significant fibrosis, and NASH were 0.86, 0.67, 0.76, and 0.66, respectively.

[0158] For the test cohort of Jiangsu Provincial People's Hospital, the AUROCs of the simplified model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH were 0.84, 0.79, 0.54, 0.76, and 0.64, respectively. The above results verify the effectiveness and wide applicability of our machine learning model based on ALT, HOMA-IR, and Mono.ratio.

[0159] As Figure 6 shown, a standard schematic diagram for identifying individuals with steatosis (controlled attenuation parameter (CAP) ≥ 274) and significant fibrosis (liver stiffness measurement (LSM) ≥ 8 kPa) in the NHANES cohort. The ROC curves and confusion matrices show the diagnostic performance of the combined model for steatosis (B) and significant fibrosis (C) in 1794 individuals serologically confirmed to be uninfected with hepatitis virus. The ROC curves and confusion matrices show the diagnostic efficacy of the combined model for steatosis (D) and significant fibrosis (E) in the NHANES cohort of 4299 individuals without missing data. TPR: True positive rate; FPR: False positive rate.

[0160] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made by those skilled in the art without creative efforts within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating various liver pathologies based on machine learning, characterized in that, It includes the following steps: Obtain the weight information, clinical examination information, and pathological evaluation results of liver tissue samples of multiple historical patients; Based on the obtained information, perform data preprocessing, construct a training set and a test set, and determine the key prediction features of each pathological index; Use the training set to train a machine learning model, determine the performance of the model in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH, and determine the final prediction index parameters and prediction model; Obtain the weight information, clinical examination information, and evaluation results of liver tissue samples of a patient, extract the relevant numerical values of the prediction index parameters, and process them with the determined trained prediction model for the corresponding prediction index parameters to obtain the final prediction result; For the Logistic regression model, add a regularization term to prevent overfitting, and set the loss function of the model as: ； Among them, J(w) is the loss function; N is the total number of samples; y i is the actual label of the i-th sample, and the actual label is 0 or 1; p i is the probability that the i-th sample is predicted as class 1; λ is the regularization strength, which is automatically selected by cross-validation; R(w) is the regularization term, and L1 regularization is adopted; The prediction formula of the model is: Among them: σ is the standard deviation of each index in the original data, w0 is the bias term, which is output by the model; w j is the weight, that is, the contribution of each feature to the model, which is output by the model; x j is the j-th component of the input feature vector, which is determined by the specific numerical value of each sample in the training set; If p(y = 1|x) ≥ 0.5, predict that the sample belongs to class 1; If p(y = 1|x) < 0.5, predict that the sample belongs to class 0; For the Ridge regression model, define its loss function as: where y i is the actual label of the i-th sample; w0 is the bias term of the model; w j is the weight parameter of the j-th feature; x ij is the j-th feature value of the i-th sample; α is the regularization strength; The prediction formula of the model is: where, w j is the weight coefficient of the j-th feature, output by the model; x ij is the j-th feature value of the i-th sample, determined by the specific numerical value of the index of each sample in the training set; If ≥ 0.5, predict that the class of this sample is 1; If < 0.5, predict that the category of the sample is 0; The finally selected prediction indicators include: For ballooning degeneration, select alanine aminotransferase, total bile acid, and homeostasis model assessment of insulin resistance as prediction indicators, and select the Ridge regression model; For steatosis, select alanine aminotransferase and glutamate dehydrogenase as prediction indicators, and select the Ridge regression model; For lobular inflammation, select alanine aminotransferase, prothrombin activity, and cadmium as prediction indicators, and select the Logistic regression model; For significant fibrosis, select the proportion of peripheral blood monocytes, aspartate aminotransferase, FPG, glycated albumin index, and β2-microglobulin as prediction indicators, and select the Logistic regression model; For NASH, select phospholipase A2, Cd, magnesium, fasting insulin, HOMA-IR, and adipose tissue insulin resistance index as prediction indicators, and select the Logistic regression model; For NAS score, select ALT, cobalt, triglyceride glucose body mass index as prediction indicators, and select the Logistic regression model; For risk NASH, select ALT, GA.index, and Beta2 as prediction indicators, and select the Logistic regression model.

2. The method for evaluating multiple liver pathologies based on machine learning according to claim 1, characterized in that, The weight information includes BMI data; the clinical examination information includes insulin resistance index, fasting blood glucose or glycated hemoglobin, and triglycerides; The pathological evaluation results of the liver tissue samples include ballooning degeneration, steatosis, lobular inflammation of the liver, significant fibrosis, NASH, NAS score, and risk NASH.

3. The method for evaluating multiple liver pathologies based on machine learning according to claim 1, characterized in that, The process of performing data preprocessing includes: In the training set, perform z-score transformation on the original data according to the mean and standard deviation of each index to achieve data standardization. The formula for z-score transformation is: ； Where z is the z-score value of each data, i.e., the value after standardization; x is the original value of the data; μ is the mean of each index in the original data; σ is the standard deviation of each index in the original data.

4. The method for evaluating various liver pathologies based on machine learning according to claim 1, characterized in that, During the preprocessing of the test set data, each clinical index is standardized using the mean and standard deviation of the training set, that is, subtracting the mean of the training set and dividing by the standard deviation of the training set as the standardized value to achieve the standardization of the test data; During the data preprocessing, the samples with each pathological outcome are marked as 1, and the samples without the outcome are marked as 0 for input into the machine learning model; In the training set, the best predictive features for each liver pathological outcome are screened through recursive feature elimination.

5. The method for multiple liver pathological evaluations based on machine learning according to claim 1, characterized in that Machine learning models for ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, NASH, NAS score, and risk NASH are established in sequence, and the predicted class of each sample by the model, i.e., 0 or 1, is output; Under repeated stratified 5-fold cross-validation, the balanced accuracy, precision, recall, F1 score, and AUROC of each model are calculated; The calculation methods for each index are as follows: Balanced accuracy = (1 / 2) * [(TP / (TP + FN)) + (TN / (TN + FP))]; Precision = TP / (TP + FP); Recall = TP / (TP + FN); F1 score = 2 * (Precision * Recall) / (Precision + Recall); AUROC = ∫(curve of TPR against FPR); Where TP is the true positive, TN is the true negative, FN is the false negative, and FP is the false positive.

6. The method for evaluating multiple liver pathologies based on machine learning according to claim 1, characterized in that, From the 7 independent models for each pathological result, ALT, HOMA-IR, and Mono.ratio are selected as key clinical indicators to construct a Logistic regression model with the simultaneous prediction of multiple liver pathological outcomes as the prediction target.

7. A multiple liver pathology evaluation system based on machine learning, based on the multiple liver pathology evaluation method according to any one of claims 1-6, characterized in that, Including: A data acquisition module for acquiring the weight information, clinical examination information, and evaluation results of liver tissue samples of multiple historical patients; A preprocessing module for performing data preprocessing based on the acquired information to construct a training set and a test set; A model training module for training Logistic regression and Ridge regression models using the training set, determining the performance of the models in predicting ballooning degeneration, steatosis, lobular inflammation, significant fibrosis, and NASH, and determining the final prediction index parameters and prediction models; A prediction module for acquiring the weight information, clinical examination information, and evaluation results of liver tissue samples of a patient, extracting the numerical values related to the prediction index parameters therein, and processing them with the determined trained prediction model for the corresponding prediction index parameters to obtain the final prediction result.

Images