A method and system for screening a serum marker feature of prostatic hyperplasia

Abstract

The application discloses a screening method and system for serum markers of prostatic hyperplasia, and the method comprises the following steps: first layer screening according to whether the index change amount of the subjects' physiological indexes and all serum marker indexes exceeds a set value; second layer screening of the candidate indexes obtained after the first layer screening according to the correlation coefficient with the prostate volume and the probability of prostatic hyperplasia; and finally, using a multiple logistic regression model to calculate the fitted prostate volume under different candidate index combinations, and selecting the candidate index combination with the minimum difference between the fitted prostate volume and the real prostate volume as the final screening prostatic hyperplasia detection feature. The application aims to accurately screen the feature combination with stronger correlation with prostatic hyperplasia and better accuracy for predicting prostatic hyperplasia from the subjects' physiological indexes and all serum marker indexes.

Description

Technical Field

[0001] This invention relates to a serum biomarker characterization technology for benign prostatic hyperplasia (BPH), specifically to a method and system for screening serum biomarkers for BPH. Background Technology

[0002] Benign prostatic hyperplasia (BPH) is a common disease among middle-aged and elderly men, and timely prediction and diagnosis are crucial for its treatment and prognosis. Currently, prostate-specific antigen (PSA) is a common serum biomarker used for the diagnosis of prostate diseases, but its use alone has limitations in predicting BPH. Therefore, how to achieve characteristic screening of serum biomarkers for BPH has become a key technical problem that urgently needs to be solved in the prediction of BPH. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method and system for screening serum biomarkers for benign prostatic hyperplasia (BPH) in view of the above-mentioned problems in the prior art. The present invention aims to accurately screen out the combination of features that are more strongly correlated with prostate enlargement and have better accuracy in predicting prostate enlargement from the physiological indicators of the subjects and all serum biomarker indicators.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for screening serum biomarkers for benign prostatic hyperplasia, comprising: S101, collect blood samples and prostate volume from the subject; S102, Detection of serum biomarker indicators from blood samples; S103, for the subjects' physiological indicators and all serum marker indicators, the subjects' physiological indicators and serum marker indicators are screened in the first layer based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds the set value. S104, For the candidate indicators obtained after the first layer of screening, calculate their correlation coefficients with prostate volume. The probability of prostate enlargement was calculated using a univariate logistic regression model relating prostate volume to prostate size. Based on the correlation coefficient with prostate volume and the probability of benign prostatic hyperplasia A second layer of screening is performed on the candidate indicators obtained after the first layer of screening. S105. For the candidate indicators obtained after the second layer of screening, the prostate volume fitted under different combinations of candidate indicators is calculated using a multivariate logistic regression model. The candidate indicator combination with the smallest difference between the fitted prostate volume and the actual prostate volume is selected as the final screened prostate enlargement detection feature.

[0005] Optionally, before performing the first-level screening of the subject's physiological indicators and serum marker indicators based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds a set value in step S103, the method further includes preprocessing the subject's physiological indicators and all serum marker indicators. The preprocessing includes normalization and outlier removal. The outlier refers to the indicator value whose difference between its size and mean exceeds a set value.

[0006] Optionally, in step S104, the correlation coefficient between the coefficient and prostate volume is calculated. The function expression is: , In the above formula, The correlation coefficient, The difference in rank between the i-th indicator data pairs is the difference between the rank values ​​of the i-th indicator data pairs. The serum biomarker data pairs consist of prostate volume and the subject's physiological indicators or serum biomarker indicators, where n is the total number of samples.

[0007] Optionally, in step S104, the probability of prostate enlargement is calculated using a univariate logistic regression model between the probability and the prostate volume. At that time, the functional expression of the single-factor logistic regression model used is: , In the above formula, This represents the probability of prostate enlargement. and The constant coefficients, Let i be the i-th candidate indicator.

[0008] Optionally, in step S104, based on the correlation coefficient with prostate volume... and the probability of benign prostatic hyperplasia The second layer of screening for candidate indicators obtained after the first layer of screening refers to selecting those that meet the correlation coefficient with prostate volume. Greater than the first set value, probability of prostate enlargement Candidate indicators that are greater than the second set value are used as candidate indicators after the second layer of screening.

[0009] Optionally, in step S105, when calculating the prostate volume fitted under different combinations of candidate indicators using a multiple logistic regression model, the functional expression of the multiple logistic regression model used is as follows: , In the above formula, This refers to the prostate volume. ~ The constant coefficients, ~ These are the n candidate indicators under the current candidate indicator combination.

[0010] Optionally, the candidate indicators obtained after the first layer of screening in step S103 include three physiological indicators: age, systolic blood pressure, and diastolic blood pressure, as well as five serum biomarkers: fPSA, HbA1c, IL-8, nesfatin-1, tPSA, and TG; the candidate indicators obtained after the second layer of screening in step S104 include age, IL-8, nesfatin-1, fPSA, and tPSA; and the prostate enlargement detection features finally selected in step S105 include age, IL-8, nesfatin-1, and fPSA.

[0011] Optionally, step S105 may be followed by: S201, collect the subject's age, IL-8, nesfatin-1 and fPSA; S202, the subjects' age, IL-8, nesfatin-1 and fPSA were normalized; S203, normalized age, IL-8, nesfatin-1, and fPSA are used to predict the risk coefficient of benign prostatic hyperplasia using a machine learning model. The machine learning model prediction establishes a mapping relationship between normalized age, IL-8, nesfatin-1, fPSA, and the risk coefficient of benign prostatic hyperplasia through pre-training.

[0012] The present invention also provides a serum biomarker characterization system for benign prostatic hyperplasia (BPH), comprising a microprocessor and a memory interconnected thereto, the microprocessor being programmed or configured to perform the serum biomarker characterization method for BPH. The present invention also provides a computer-readable storage medium storing a computer program for being programmed or configured by a microprocessor to perform the serum biomarker characterization method for BPH.

[0013] Compared with existing technologies, the present invention has the following advantages: The method of the present invention includes a first-level screening of candidate indicators based on the physiological indicators and all serum biomarkers of the subjects, according to whether the change in the indicator of whether or not prostate hyperplasia exists exceeds a set value. The correlation between candidate indicators and the correlation between candidate indicators and prostate volume are calculated. Based on the correlation, the predictive efficacy of a single serum biomarker indicator is analyzed by univariate power analysis. Candidate indicators whose predictive efficacy indicators in the univariate analysis are greater than the set value are included in a multivariate logistic regression model to be used to predict the quantitative indicators of prostate enlargement using the multivariate logistic regression model. Thus, the two-level screening, combined with univariate analysis and multivariate logistic regression model, can accurately screen out feature combinations that are more strongly correlated with prostate enlargement and have better accuracy in predicting prostate enlargement from the subjects' physiological indicators and all serum biomarkers. This effectively solves the limitations of using prostate-specific antigen (PSA) alone as a serum biomarker for predicting prostate hyperplasia in existing technologies. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention. Detailed Implementation

[0015] like Figure 1 As shown, the serum biomarker screening method for benign prostatic hyperplasia in this embodiment includes: S101, collect blood samples and prostate volume from the subject; S102, Detection of serum biomarker indicators from blood samples; S103, for the subjects' physiological indicators and all serum marker indicators, the subjects' physiological indicators and serum marker indicators are screened in the first layer based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds the set value. S104, For the candidate indicators obtained after the first layer of screening, calculate their correlation coefficients with prostate volume. The probability of prostate enlargement was calculated using a univariate logistic regression model relating prostate volume to prostate size. Based on the correlation coefficient with prostate volume and the probability of benign prostatic hyperplasia A second layer of screening is performed on the candidate indicators obtained after the first layer of screening. S105. For the candidate indicators obtained after the second layer of screening, the prostate volume fitted under different combinations of candidate indicators is calculated using a multivariate logistic regression model. The candidate indicator combination with the smallest difference between the fitted prostate volume and the actual prostate volume is selected as the final screened prostate enlargement detection feature.

[0016] In step S101 of this embodiment, when collecting blood samples and prostate volume from the subjects, the prostate volume is also graded to obtain a volume grading, which is also used as a physiological indicator. In this embodiment, the subjects included 114 men with benign prostatic hyperplasia and an age-matched healthy control group (66 people).

[0017] In step S102 of this embodiment, when detecting serum biomarkers from a blood sample, the detected serum biomarkers include: IL-8 (interleukin-8), insulin, nesfatin-1 (a derivative of nucleoside 2), GLU (glucose glucose), HOMA-IR (insulin resistance index), HbA1c (glycated hemoglobin), fPSA (free prostate-specific antigen), fPSA / tPSA (a percentage of total prostate-specific antigen), uric acid, TG (triglycerides), TC (total cholesterol), LL (low-density lipoprotein), HL (high-density lipoprotein), and creatinine. It should be noted that the above serum biomarkers are all well-known indicators, and their detection methods are all existing methods; therefore, their specific components and detection methods will not be detailed here.

[0018] In step S103 of this embodiment, the physiological indicators of the subjects include age, systolic blood pressure, diastolic blood pressure, BMI, waist circumference, diabetes, hypertension, and hyperlipidemia. When performing the first-level screening of the subjects' physiological indicators and serum biomarkers based on whether the changes in indicators related to the presence or absence of benign prostatic hyperplasia exceed set values, independent samples t-tests and Wilcoxon independent trigonometric identities can be used to analyze the differences in various parameters in the serum of individuals with benign prostatic hyperplasia and healthy individuals. Finally, the candidate indicators obtained after the first-level screening in step S103 include three physiological indicators: age, systolic blood pressure, and diastolic blood pressure, as well as five serum biomarkers: fPSA, HbA1c, IL-8, nesfatin-1, tPSA, and TG.

[0019] In this embodiment, before performing the first-level screening of the subject's physiological indicators and serum biomarkers based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds a set value in step S103, the method further includes preprocessing the subject's physiological indicators and all serum biomarkers. The preprocessing includes normalization and outlier removal. Outliers refer to indicator values ​​whose difference between their size and mean exceeds a set value.

[0020] In step S104 of this embodiment, the correlation coefficient between the coefficient and the prostate volume is calculated. The function expression is: , In the above formula, The correlation coefficient, The difference in rank between the i-th indicator data pairs is the difference between the rank values ​​of the i-th indicator data pairs. The serum biomarker data pairs consist of prostate volume and the subject's physiological indicators or serum biomarker indicators, where n is the total number of samples.

[0021] In step S104 of this embodiment, the probability of prostate enlargement is calculated using a single-factor logistic regression model between the probability and prostate volume. At that time, the functional expression of the single-factor logistic regression model used is: , In the above formula, This represents the probability of prostate enlargement. and The constant coefficients, Let be the i-th candidate indicator. It should be noted that the constant coefficients for different constants are all different. For example, in this embodiment, the function expression of the univariate logistic regression model for the five serum biomarker characteristics is: , , , , , , In the above formulas, the last parameter is the serum biomarker index.

[0022] In step S104 of this embodiment, the correlation coefficient with prostate volume is used... and the probability of benign prostatic hyperplasia The second layer of screening for candidate indicators obtained after the first layer of screening refers to selecting those that meet the correlation coefficient with prostate volume. The probability of prostate enlargement is greater than a first preset value (which can be set according to actual needs, for example, 0.3 in this embodiment). Candidate indicators that are greater than a second set value (which can be set according to actual needs, for example, 0.01 in this embodiment) are used as candidate indicators obtained after the second layer of screening. Finally, in step S104, for the candidate indicators obtained after the first layer of screening (including three physiological indicators: age, systolic blood pressure, and diastolic blood pressure, and five serum biomarkers: fPSA, HbA1c, IL-8, nesfatin-1, tPSA, and TG), the candidate indicators obtained after the second layer of screening include age, IL-8, Nesfatin-1, fPSA, and tPSA.

[0023] In step S105 of this embodiment, when calculating the prostate volume fitted under different combinations of candidate indicators using a multivariate logistic regression model, the functional expression of the multivariate logistic regression model used is as follows: , In the above formula, This refers to the prostate volume. ~ The constant coefficients, ~ For each of the current candidate indicator combinations, n candidate indicators are selected. Multiple linear regression is performed with prostate volume as the dependent variable Y and age, IL-8, Nesfatin-1, fPSA, and tPSA as independent variables to analyze the influence of these parameters on changes in prostate volume. Finally, the prostate enlargement detection features finally selected in step S105 include age, IL-8, nesfatin-1, and fPSA.

[0024] Furthermore, this embodiment also provides a serum biomarker characterization system for benign prostatic hyperplasia (BPH), including a microprocessor and a memory interconnected thereto, the microprocessor being programmed or configured to execute the serum biomarker characterization method for BPH. This embodiment also provides a computer-readable storage medium storing a computer program for being programmed or configured by the microprocessor to execute the serum biomarker characterization method for BPH.

[0025] Example 2: This embodiment is a further application of Embodiment 1. Specifically, this embodiment also provides a method for predicting the risk factor of benign prostatic hyperplasia (BPH) using age, IL-8, nesfatin-1, and fPSA, as a reference indicator for the diagnosis of BPH. Step S105 is followed by: S201, collect the subject's age, IL-8, nesfatin-1 and fPSA; S202, the subjects' age, IL-8, nesfatin-1 and fPSA were normalized; S203 uses a machine learning model to predict the risk coefficient of benign prostatic hyperplasia (BPH) by normalizing age, IL-8, nesfatin-1, and fPSA. The machine learning model prediction establishes a mapping relationship between normalized age, IL-8, nesfatin-1, fPSA, and the risk coefficient of BPH.

[0026] Machine learning models can employ various methods as needed, such as Support Vector Machines, Logistic Regression, Random Forest, and XGBoost. This embodiment utilizes machine learning methods (including Support Vector Machines, Logistic Regression, Random Forest, and XGBoost) to batch train biomarker data and construct a machine learning model. A five-fold cross-validation method is used to split and validate the data, using four folds as the training set and one fold as the internal validation set. The average of five iterations is used as the model's evaluation. The AUC (Area Under Curve), sensitivity, and specificity of the machine learning models obtained through cross-validation are compared among models, and the best-performing XGBoost model is selected for subsequent application and development. XGBoost is a decision tree-based machine learning algorithm where the generation of each subsequent decision tree is based on the prediction results of the previous one. Through regularization, learning rate (shrinkage), column sampling, and approximate optimal split points, it exhibits excellent performance in preventing overfitting, handling missing values, and computational efficiency. By inputting age, IL-8, nesfatin-1, and fPSA into the trained XGBoost model, the risk coefficient for benign prostatic hyperplasia can be obtained. Experimental results show that the trained XGBoost model has an AUC of 0.9992 on the training set and 0.8456 on the validation set. The calibration curve demonstrates a good fit, and the decision curve shows significant clinical benefits from using the model. It offers the following advantages: High accuracy: By combining multiple biomarkers, it can provide higher predictive accuracy; Early prediction: Risk prediction can be performed in the early stages of disease; Non-invasive: Only routine blood sampling is required, making it non-invasive and patient-friendly.

[0027] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for screening serum biomarkers for benign prostatic hyperplasia, characterized in that, include: S101, collect blood samples and prostate volume from the subject; S102, Detection of serum biomarker indicators from blood samples; S103, for the subjects' physiological indicators and all serum marker indicators, the subjects' physiological indicators and serum marker indicators are screened in the first layer based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds the set value. S104, For the candidate indicators obtained after the first layer of screening, calculate their correlation coefficients with prostate volume. The probability of prostate enlargement was calculated using a univariate logistic regression model relating it to prostate volume. Based on the correlation coefficient with prostate volume and the probability of benign prostatic hyperplasia A second layer of screening is performed on the candidate indicators obtained after the first layer of screening. S105. For the candidate indicators obtained after the second layer of screening, the prostate volume fitted under different combinations of candidate indicators is calculated using a multivariate logistic regression model. The candidate indicator combination with the smallest difference between the fitted prostate volume and the actual prostate volume is selected as the final screened prostate enlargement detection feature. The candidate indicators obtained after the first layer of screening in step S103 include three physiological indicators: age, systolic blood pressure, and diastolic blood pressure, as well as five serum biomarkers: fPSA, HbA1c, IL-8, nesfatin-1, tPSA, and TG. The candidate indicators obtained after the second layer of screening in step S104 include age, IL-8, nesfatin-1, fPSA, and tPSA. The prostate enlargement detection features finally selected in step S105 include age, IL-8, nesfatin-1, and fPSA.

2. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 1, characterized in that, Before performing the first-level screening of the subject's physiological indicators and serum marker indicators based on whether the change in the indicator of whether or not there is benign prostatic hyperplasia exceeds the set value in step S103, the method further includes preprocessing the subject's physiological indicators and all serum marker indicators. The preprocessing includes normalization and outlier removal. Outliers refer to indicator values ​​whose difference between their size and mean exceeds the set value.

3. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 2, characterized in that, In step S104, the correlation coefficient between the prostate volume and the volume of the prostate is calculated. The function expression is: ， In the above formula, The correlation coefficient, The difference in rank between the i-th indicator data pairs is the difference between the rank values ​​of the i-th indicator data pairs. The serum biomarker data pairs consist of prostate volume and the subject's physiological indicators or serum biomarker indicators, where n is the total number of samples.

4. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 3, characterized in that, In step S104, the probability of prostate enlargement is calculated using a univariate logistic regression model between the probability and prostate volume. At that time, the functional expression of the single-factor logistic regression model used is: ， In the above formula, This represents the probability of prostate enlargement. and The constant coefficients, Let i be the i-th candidate indicator.

5. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 3, characterized in that, In step S104, based on the correlation coefficient with prostate volume and the probability of benign prostatic hyperplasia The second layer of screening for candidate indicators obtained after the first layer of screening refers to selecting those that meet the correlation coefficient with prostate volume. Greater than the first set value, probability of prostate enlargement Candidate indicators that are greater than the second set value are used as candidate indicators after the second layer of screening.

6. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 3, characterized in that, In step S105, when calculating the prostate volume fitted under different combinations of candidate indicators using a multiple logistic regression model, the functional expression of the multiple logistic regression model used is as follows: ， In the above formula, This refers to the prostate volume. ~ The constant coefficients, ~ These are the n candidate indicators under the current candidate indicator combination.

7. The method for screening serum biomarkers for benign prostatic hyperplasia according to claim 1, characterized in that, Step S105 is followed by: S201, collect the subject's age, IL-8, nesfatin-1 and fPSA; S202, the subjects' age, IL-8, nesfatin-1 and fPSA were normalized; S203, normalized age, IL-8, nesfatin-1, and fPSA are used to predict the risk coefficient of benign prostatic hyperplasia using a machine learning model. The machine learning model prediction establishes a mapping relationship between normalized age, IL-8, nesfatin-1, fPSA, and the risk coefficient of benign prostatic hyperplasia through pre-training.

8. A serum biomarker characterization system for benign prostatic hyperplasia, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to perform the serum biomarker characterization method for benign prostatic hyperplasia as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that, The computer program is used to be programmed or configured by a microprocessor to perform the serum biomarker characterization method for benign prostatic hyperplasia as described in any one of claims 1 to 7.

Images