Human kidney physiological age prediction method based on machine learning

Abstract

According to the human kidney physiological age prediction method based on machine learning, five blood detection indexes of creatinine, albumin, urea, uric acid and total protein closely related to the kidney are collected in a targeted manner, and computer model modeling is performed on male and female data according to kidney health data; and constructing a computer model for predicting the physiological ages of the kidneys of the male and female human bodies to evaluate the physiological ages of the kidneys of the individual bodies with different genders. According to the human kidney physiological age prediction method based on machine learning, core data only uses data indexes of creatinine, albumin, urea, uric acid and total protein related to the human kidney for modeling, a physiological age model of the human kidney is constructed, and the physiological age of the human kidney can be evaluated in a targeted mode; moreover, according to the human kidney physiological age prediction method based on machine learning, modeling is performed on the male and the female based on the data difference of the male and the female genders, so that the prediction result is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly to a method for predicting the physiological age of the human kidney based on machine learning with good prediction effect. Background Art

[0002] The kidneys are important excretory organs of the human body, occurring in pairs and located on both sides of the lower back. Their main function is to filter the blood, excrete metabolic waste and excess water, and form urine. The kidneys are also involved in regulating blood pressure, electrolyte balance, and red blood cell production. Each kidney consists of about 1 million nephrons, including structures such as glomeruli and renal tubules. The glomeruli are responsible for filtration, while the renal tubules are responsible for reabsorption and secretion functions. The health of the kidneys is crucial for maintaining overall human health.

[0003] Creatinine, albumin, urea, uric acid, and total protein in physical examinations are important indicators for evaluating kidney function. Creatinine is produced by muscle metabolism and excreted by the kidneys. An increase in its level usually indicates a decline in glomerular filtration function. Urea is the end product of protein metabolism and is excreted by the kidneys. An increase in urea may reflect impaired kidney function or abnormal protein metabolism. Uric acid is a product of purine metabolism and is mainly excreted by the kidneys. An increase in its level may be related to kidney excretion disorders or a high-purine diet. Albumin is the main protein in plasma. If albumin is detected in urine (such as microalbuminuria), it may indicate early kidney damage, such as diabetic nephropathy. Total protein reflects the overall protein metabolism status of the body and often decreases in patients with kidney diseases due to protein loss. Abnormalities in these indicators can suggest kidney function problems and further examinations are needed to clarify the reasons.

[0004] Calendar age refers to the length of time from birth to the present, calculated in years, which is fixed and unchangeable. Physiological age reflects the actual health status and functional level of the body and may not be consistent with the calendar age. Physiological age is affected by various factors such as lifestyle, genetics, environment, and diseases. For example, a healthy diet and regular exercise can delay physiological age, while bad habits such as smoking and drinking can accelerate aging. Compared with calendar age, physiological age can better reflect an individual's true health status, is an important indicator for measuring the body state, and helps to develop personalized health management plans.

[0005] Kidney physiological age is a quantitative indicator for evaluating the degree of physiological aging of an individual's kidneys based on the functional status and health indicators of the kidneys. It reflects the actual functional age of the kidneys and can be used to measure the health status and aging rate of the kidneys. Referring to the article published by Tian, Y.E et al. in "Nature Medicine" in 2023, the physiological age of disease patients is significantly higher than that of healthy individuals, and bad living habits (such as smoking, drinking, staying up late, etc.) are also associated with a higher physiological age.

[0006] Most of the existing technical solutions use all physical examination indicators to construct machine learning models, and do not divide relevant indicators into corresponding organs or systems. However, different organs or systems of an individual often have different physiological age indexes. A single overall model is difficult to accurately reflect the physiological ages of different organs or systems of an individual. Moreover, there are significant differences in some indicators between men and women, and a single model is difficult to reflect the physiological ages of different organs or systems of individuals of different genders.

[0007] Therefore, it is necessary to propose an improvement to overcome the defects of the existing technology. Summary of the Invention

[0008] The purpose of the present invention is to solve the problems in the existing technology and provide a machine learning-based method for predicting the physiological age of the human kidney with good prediction effect.

[0009] The technical solution of the present invention is as follows: A method for predicting the physiological age of the human kidney based on machine learning, comprising the following steps: S1. Collect physical examination data of the human kidney: the physical examination data of the human kidney includes the gender of the person being examined, calendar age, and data of creatinine, albumin, urea, uric acid, and total protein; S2. Divide the health examination data into a male data set and a female data set according to gender; S3. Divide the male data set and the female data set into a training set and a test set according to a ratio respectively; S4. Perform normalization processing on the data of the training set: perform normalization processing on the data in the training sets of the male data set and the female data set respectively; S5. Build a model and perform model training: build models for the data in the training sets of the male data set and the female data set respectively, and perform cross-validation hyperparameter optimization by means of grid search; S6. Perform correction on the prediction of the physiological age of the human kidney: perform age correction according to the regression to the mean effect, use the calendar age as the independent variable and the original age difference as the dependent variable for linear regression; original age difference = slope * calendar age + intercept + residual; wherein, the residual is the age difference after correction, that is, the part related to the calendar age is removed from the original age difference, and the formula is: corrected age difference = original age difference – (slope * original age difference + intercept); wherein, the slope and the intercept are calculated, and the specific calculation method is: perform linear fitting on the original age difference and the calendar age in the training sets of the male data set and the female data set respectively to obtain the slope and the intercept in the male data set and the female data set.

[0010] As a preferred technical solution, after step S6, the following steps are further included: S7. Calculate the senescence index: the formula for calculating the senescence index is: senescence index = corrected age difference / calendar age.

[0011] As a preferred technical solution, the ratio of the data volume of the training set to the test set in step S3 is 2:1, 3:1, 4:1 or 5:1.

[0012] As a preferred technical solution, in step S4 for normalizing the data of the training set, the normalization method can be any one of Z-score standardization, maximum-minimum normalization, median normalization, and maximum absolute value normalization.

[0013] As a further preferred technical solution, in step S4 for normalizing the data of the training set, the normalization method is maximum absolute value normalization, and the formula is x′ = x / max(|x|), where x is the original value and max(|x|) is the maximum value of the absolute values of all values in the dataset.

[0014] As a preferred technical solution, in step S5, the following one or more regression machine learning models are used for modeling: SVR, LightGBM, CatBoost, Ridge, RandomForest, XGBoost, Huber, Lasso, ElasticNet.

[0015] As a further preferred technical solution, in step S5, SVR is used for modeling; in step S5, the data of the training sets in the male dataset and the female dataset are respectively modeled, and the hyperparameter optimization by cross-validation in the way of grid search specifically includes the following steps: S51, create a model; S52, define a parameter grid: the parameter grid includes a penalty coefficient, a kernel function, a kernel function coefficient, and a slack variable; S53, perform hyperparameter tuning; S54, obtain the best parameters and model.

[0016] As a further preferred technical solution, after step S54, the following steps are further included: S55, perform evaluation and verification on the training set and / or the test set.

[0017] As another further preferred technical solution, step S53 for hyperparameter tuning is specifically to use GridSearchCV for hyperparameter tuning.

[0018] A method for predicting the physiological age of the human kidney based on machine learning of the present invention collects 5 blood test indexes closely related to the kidney, namely creatinine, albumin, urea, uric acid and total protein, and respectively models computer models for male and female data around the kidney health data, constructs a computer model for predicting the physiological age of the human kidney of men and women, and is used to evaluate the physiological age of the human kidney of individuals of different genders. The method for predicting the physiological age of the human kidney based on machine learning of the present invention constructs a physiological age model of the human kidney by only using data indexes such as creatinine, albumin, urea, uric acid and total protein related to the human kidney as core data for modeling, and can specifically evaluate the physiological age of the human kidney; moreover, the method for predicting the physiological age of the human kidney based on machine learning of the present invention respectively models for men and women based on the data differences between men and women, making the prediction results more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flowchart of the specific implementation manner of a method for predicting the physiological age of the human kidney based on machine learning of the present invention; Figure 2 is the prediction result of the male kidney physiological age model in the test set in a method for predicting the physiological age of the human kidney based on machine learning of this embodiment; Figure 3 is the prediction result of the female kidney physiological age model in the test set in a method for predicting the physiological age of the human kidney based on machine learning of this embodiment; Figure 4 is the male age difference after age prediction correction through step S6 in a method for predicting the physiological age of the human kidney based on machine learning of this embodiment; Figure 5 is the female age difference after age prediction correction through step S6 in a method for predicting the physiological age of the human kidney based on machine learning of this embodiment. SPECIFIC IMPLEMENTATION MANNER

[0020] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. "Multiple" generally includes at least two, but does not exclude the case of including at least one.

[0022] It should be understood that the term "and / or" used herein is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0023] Depending on the context, the words "if", "when" as used herein can be interpreted as "when...", "when...", "in response to determining", or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined", "in response to determining", "when detecting (stated condition or event)", or "in response to detecting (stated condition or event)".

[0024] It should also be noted that the term "comprising", "including", or any other variant thereof is intended to cover non-exclusive inclusion, such that a commodity or system comprising a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such commodity or system. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the commodity or system comprising said element.

[0025] As Figure 1 shown is a specific implementation manner of a method for predicting the physiological age of the human kidney based on machine learning according to the present invention. A method for predicting the physiological age of the human kidney based on machine learning in this embodiment includes the following steps: S1. Collect human kidney physical examination data: The human kidney physical examination data includes the gender of the person being examined, calendar age, and data on creatinine, albumin, urea, uric acid, and total protein; S2. Divide the health examination data into a male data set and a female data set according to gender; S3. Divide the male data set and the female data set into a training set and a test set according to a proportion respectively; S4. Perform normalization processing on the data of the training set: Perform normalization processing on the data in the training sets of the male data set and the female data set respectively; S5. Modeling and model training: Model the data in the training sets of the male dataset and the female dataset respectively, and perform hyperparameter optimization through cross-validation using grid search. S6. Human kidney physiological age prediction and correction: Perform age correction according to the regression to the mean effect. Use the calendar age as the independent variable and the original age difference as the dependent variable for linear regression. Original age difference = slope * calendar age + intercept + residual; Among them, the residual is the age difference after correction, that is, the part related to the calendar age is removed from the original age difference. The formula is: Corrected age difference = original age difference – (slope * original age difference + intercept); Among them, the slope and intercept are calculated. The specific calculation method is: linearly fit the original age difference and the calendar age in the training sets of the male dataset and the female dataset respectively to obtain the slope and intercept in the male dataset and the female dataset.

[0026] In this embodiment, after step S6, the following steps are further included: S7. Calculate the aging index: The formula for calculating the aging index is: Aging index = corrected age difference / calendar age.

[0027] Specifically, for a method for predicting the physiological age of the human kidney based on machine learning in this implementation, 1403 male individual samples were collected, and the age distribution range was 15 - 90 years old; 1544 female individual samples were collected, and the age distribution range was 17 - 80 years old.

[0028] In step S3, the ratio of the data volume of the training set to the test set is 3:1. After division: Among male individuals, the number of training set samples is 1053, and the number of test set samples is 350; among female individuals, the number of training set samples is 1185, and the number of test set samples is 386. It should be known that in practical applications, the ratio of the training set to the test set can be 4:1, 3:1, 2:1, or 5:1 according to needs, and it will not affect the implementation of the solution of the present invention.

[0029] In practical applications, in step S4 for normalizing the data of the training set, the normalization method can be any one of Z-score standardization, maximum-minimum normalization, median normalization, and maximum absolute value normalization. For a method for predicting the physiological age of the human kidney based on machine learning in this embodiment, in step S4 for normalizing the data of the training set, the maximum absolute value normalization method is selected. The formula is x′ = x / max(|x|), where x is the original value and max(|x|) is the maximum value of the absolute values of all values in the dataset.

[0030] A method for predicting the physiological age of the human kidney based on machine learning. In specific applications, in step S5, the following one or more regression machine learning models can be used for modeling: SVR, LightGBM, CatBoost, Ridge, RandomForest, XGBoost, Huber, Lasso, ElasticNet. In a method for predicting the physiological age of the human kidney based on machine learning in this embodiment, in step S5, SVR is used to build a regression machine learning model. In a method for predicting the physiological age of the human kidney based on machine learning in this embodiment, in step S5, the data in the training sets of the male dataset and the female dataset are respectively modeled, and hyperparameter optimization is performed by means of grid search, which specifically includes the following steps: S51. Create a model; S52. Define a parameter grid: The parameter grid includes a penalty coefficient, a kernel function, a kernel function coefficient, and a slack variable; S53. Perform hyperparameter tuning; S54. Obtain the optimal parameters and model.

[0031] A method for predicting the physiological age of the human kidney based on machine learning. In actual applications, after step S54, the following steps are further included: S55. Perform evaluation and verification on the training set and / or the test set. In a method for predicting the physiological age of the human kidney based on machine learning in this embodiment, step S55 is to perform evaluation and verification on both the training set and the test set. And in this embodiment, in step S53, hyperparameter tuning is specifically performed using GridSearchCV.

[0032] The specific implementation method of step S5 of a method for predicting the physiological age of the human kidney based on machine learning in this embodiment, using male data, maximum absolute value normalization, and SVR modeling in Python is as follows: Import the necessary Python packages: import numpy as np import pandas as pd from sklearn.model_selection import train_test_split, GridSearchCV from sklearn.pipeline import Pipeline from sklearn.preprocessing import StandardScaler, MaxAbsScaler from sklearn.svm import SVR from sklearn.metrics import mean_squared_error from scipy.stats import pearsonr # Read in the data for modeling: data = pd.read_csv(“Kidney Physiological Age Modeling Data.csv”, header =0, index_col= 0) # Obtain male data maledf = data[data[“gender”]==”male”] # Obtain the matrix of test values for 5 indicators (UREA: urea; SCR: creatinine; TP: total protein; ALB: albumin; UA: uric acid): X = maledf[['UREA','SCR','TP','ALB','UA']] # Obtain the age y = maledf["age"] # Split the training set and the test set X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0) # Create a modeling Pipeline: pipeline = Pipeline( ("scaler", MaxAbsScaler()), ("svm", SVR()) ) # Define the parameter grid: param_grid = { "svm__C": [0.1, 1, 10, 100], # Penalty coefficient "svm__kernel": ["linear", "rbf"], # Kernel function "svm__gamma": ["scale", "auto"], # Kernel function coefficient "svm__epsilon": [0.1, 0.2, 0.5], # Slack variable } # Hyperparameter Tuning Using GridSearchCV: grid_search = GridSearchCV( estimator=pipeline, param_grid=param_grid, cv=5, scoring="neg_mean_squared_error", verbose=1 ) grid_search.fit(X_train, y_train) # Obtain the Best Parameters and Model: best_params = grid_search.best_params_ best_model = grid_search.best_estimator_ # Evaluate on the Training Set y_train_pred = best_model.predict(X_train) train_mae = mean_squared_error(y_train, y_train_pred) train_r, train_p = pearsonr(y_train, y_train_pred) # Evaluate on the Test Set: y_test_pred = best_model.predict(X_test) test_mae = mean_squared_error(y_test, y_test_pred) test_r, test_p = pearsonr(y_test, y_test_pred) A method for predicting the physiological age of the human kidney based on machine learning in this embodiment. In step S6, the slope and intercept are calculated. The specific calculation method is as follows: linearly fit the original age difference and the calendar age in the training sets of the male dataset and the female dataset respectively to obtain the slope and intercept in the male dataset and the female dataset. In this embodiment, when the normalization method is the maximum absolute value normalization and the modeling uses SVR, by linearly fitting the original age difference and the calendar age in the training sets of the male dataset and the female dataset, the slope and intercept in the male dataset and the female dataset are obtained as follows: the male slope is -0.8796, the male intercept is 46.72, the female slope is -0.9289, and the female intercept is 46.97.

[0033] A method for predicting the physiological age of the human kidney based on machine learning in this embodiment has achieved good performance in practical applications. In the test set, as Figure 2 shown, the Pearson correlation coefficient between the physiological age predicted by the male cardiovascular physiological age model and the calendar age has reached 0.27. As Figure 3 shown, the correlation coefficient between the female physiological age predicted by the female cardiovascular physiological age model and the calendar age has reached 0.21, both showing relatively good modeling effects. However, both models have the "regression to the mean effect" in statistics, that is, for young individuals, the predicted biological age is often higher than the calendar age; for older individuals, the predicted biological age is often lower than the calendar age. After age prediction correction through step S6, the age differences after correction for males and females are respectively as Figure 4 and Figure 5 shown. The MAE (Mean Absolute Error) after correction for males is 3.10, and the MAE after correction for females is 2.21, achieving very good results.

[0034] A method for predicting the physiological age of the human kidney based on machine learning in this embodiment also defines and calculates the aging index through step S7. The aging index represents the number of years an individual ages each year and can evaluate the aging rate. If the aging index is greater than 1, it means that the individual's aging speed is higher than the normal speed and the aging speed is faster; if the aging index is less than 1, it represents that the individual's aging speed is slower than the normal speed and the aging speed is slower.

[0035] A method for predicting the physiological age of the human kidney based on machine learning collects five blood test indicators closely related to the kidney, namely creatinine, albumin, urea, uric acid, and total protein, in a targeted manner. Computer models are built for male and female data respectively around the kidney health data to construct a computer model for predicting the physiological age of the human kidney of different genders, so as to evaluate the physiological age of the human kidney of individuals of different genders. In the method for predicting the physiological age of the human kidney based on machine learning of the present invention, since only the data indicators of creatinine, albumin, urea, uric acid, and total protein related to the human kidney are used as the core data for modeling to construct the physiological age model of the human kidney, the physiological age of the human kidney can be evaluated in a targeted manner; moreover, in the method for predicting the physiological age of the human kidney based on machine learning of the present invention, modeling is carried out separately for men and women based on the data differences between genders, making the prediction results more accurate.

[0036] The above is only a preferred embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. That is, all equivalent changes and modifications made to the content of the scope of the patent application of the present invention shall fall within the technical scope of the present invention.

Claims

1. A method for predicting the physiological age of the human kidney based on machine learning, characterized in that: It includes the following steps: S1. Collect human kidney physical examination data: The human kidney physical examination data includes the gender of the examinee, calendar age, and data on creatinine, albumin, urea, uric acid, and total protein; S2. Divide the health examination data into a male data set and a female data set according to gender; S3. Divide the male data set and the female data set into a training set and a test set respectively in proportion; S4. Perform normalization on the data of the training set: Perform normalization on the data in the training sets of the male data set and the female data set respectively; S5. Build a model and conduct model training: Build models for the data in the training sets of the male data set and the female data set respectively, and perform cross-validation hyperparameter optimization in the way of grid search; S6. Perform prediction and correction of the physiological age of the human kidney: Perform age correction according to the regression to the mean effect, use the calendar age as the independent variable, and the original age difference as the dependent variable for linear regression; Original age difference = slope * calendar age + intercept + residual; Among them, the residual is the age difference after correction, that is, the part related to the calendar age is removed from the original age difference, and the formula is: Corrected age difference = original age difference – (slope * original age difference + intercept); Among them, the slope and intercept are calculated. The specific calculation method is: Perform linear fitting of the original age difference and the calendar age in the training sets of the male data set and the female data set respectively to obtain the slope and intercept in the male data set and the female data set.

2. The method for predicting the physiological age of the human kidney based on machine learning according to claim 1, wherein: After step S6, the following steps are also included: S7. Calculate the aging index: The formula for calculating the aging index is: Aging index = corrected age difference / calendar age.

3. A method for predicting the physiological age of the human kidney based on machine learning according to claim 1, characterized in that: In step S3, the ratio of the data volume of the training set to the test set is 2:1, 3:1, 4:1, or 5:

1.

4. A method for predicting the physiological age of the human kidney based on machine learning according to claim 1, characterized in that: In step S4, when performing normalization on the data of the training set, the normalization method can be any one of Z-score standardization, maximum-minimum normalization, median normalization, and maximum absolute value normalization.

5. A method for predicting the physiological age of the human kidney based on machine learning according to claim 4, characterized in that: In step S4, when performing normalization on the data of the training set, the normalization method is maximum absolute value normalization, and the formula is x′ = x / max(|x|), where x is the original value, and max(|x|) is the maximum value of the absolute values of all values in the data set.

6. A method for predicting the physiological age of the human kidney based on machine learning according to claim 1, characterized in that: In step S5, the following one or more regression machine learning models are used for modeling: SVR, LightGBM, CatBoost, Ridge, RandomForest, XGBoost, Huber, Lasso, ElasticNet.

7. A method for predicting the physiological age of the human kidney based on machine learning according to claim 6, characterized in that: In step S5, SVR modeling is used; In step S5, models are built for the data in the training sets of the male data set and the female data set respectively, and the specific steps for performing cross-validation hyperparameter optimization in the way of grid search include the following steps: S51. Create a model; S52. Define the parameter grid: The parameter grid includes the penalty coefficient, kernel function, kernel function coefficient, and relaxation variable; S53. Perform hyperparameter tuning; S54. Obtain the best parameters and model.

8. A method for predicting the physiological age of the human kidney based on machine learning according to claim 7, characterized in that: After step S54, the following steps are also included: S55. Perform evaluation and verification on the training set and / or test set.

9. A method for predicting the physiological age of the human kidney based on machine learning according to claim 7, characterized in that: Step S53 for hyperparameter tuning specifically uses GridSearchCV for hyperparameter tuning.

Images