Prostate cancer three-classification risk layering method based on integrated learning model

Abstract

The invention discloses a prostate cancer three-classification risk layering method based on an integrated learning model, and relates to the technical field of data processing and analysis. The method comprises the following steps: collecting clinical information and pathological data of a patient with increased PSA, and dividing the data into a training set and a test set; the training set is preprocessed, and prediction features are screened through LASSO regression; constructing a plurality of machine learning base models based on the features, and training and optimizing through cross validation; soft voting is constructed through an integration strategy, and an integration model is stacked; setting double thresholds according to the integrated model prediction probability, and establishing a layering rule; combining an integrated model and rules to form a three-classification model, and judging low, high and medium risks according to probabilities; and finally verifying the model diagnosis performance in the test set. According to the method, through cross-modal feature integration and ensemble learning, the method is PSAlt; accurate risk stratification is provided for 30 ng / mL people, biopsy decision-making efficiency is optimized, and excessive puncture and missed diagnosis risks are reduced.

Claims

1. A three-category risk stratification method for prostate cancer based on an ensemble learning model, characterized in that: The steps include: S1 collects clinical information and pathological results of patients with elevated PSA to form a data set, which is divided into a training set and a test set; S2 preprocesses the training set data, uses LASSO regression to screen clinical information variables, and retains the prediction features that meet the requirements; Based on the filtered features, S3 builds multiple different types of machine learning base models on the training set and uses cross-validation methods to train and optimize the parameters of each base model; S4 combines multiple base models through an ensemble learning strategy to construct a soft voting ensemble model and a stacking ensemble model, respectively, to obtain a binary classification ensemble model that outputs prediction probabilities; S5 sets a double-threshold cutoff point based on the prediction probability of the integrated model, including a first threshold and a second threshold, and establishes a double-threshold stratification rule; S6 combines the integrated model with the double-threshold stratification rule to form a complete three-category risk stratification model. When the predicted probability is lower than the first threshold, it is judged as low risk; when it is higher than the second threshold, it is judged as high risk; when it is between the two thresholds, it is judged as medium risk; S7 verifies the diagnostic performance of the three-category risk stratification model on the test set.

2. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: Clinical information includes hematological indicators, imaging indicators and clinical characteristics. Hematological indicators include prostate-specific antigen, free prostate-specific antigen, prostate health index, neutrophils, lymphocytes, prostate-specific antigen precursor 2, neutrophil-to-lymphocyte ratio, and platelet-to-lymphocyte ratio. Imaging indicators include Prostate Imaging Reporting and Data System score and prostate volume. Clinical characteristics include age, history of prostate puncture, and family history.

3. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: LASSO regression variables were screened and finally retained 10 predictive features, including age, prostate imaging reporting and data system score, prostate health index, prostate health index density, prostate specific antigen density, prostate specific antigen precursor 2, lymphocytes, neutrophils, urine white blood cells, and prostate volume.

4. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: The machine learning base models include the extreme gradient boosting tree model based on gradient boosting, the category boosting model based on gradient boosting, and the random forest model based on classification and regression trees. The hyperparameters of each base model are optimized by grid search combined with five-fold cross validation.

5. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: The soft voting ensemble model achieves classification by weighted averaging the prediction probabilities of each base model, and the weights are dynamically determined based on the area under the receiver operating characteristic curve of each base model on the validation set. The stacked ensemble model adopts a two-stage training architecture. In the first stage, each base model performs five-fold cross-training on the training set to generate meta-features. In the second stage, the logistic regression model is used as a meta-learner to train based on the meta-features to output the final prediction results.

6. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: The dual threshold determination method is to traverse the prediction probability thresholds of the integrated model, select the first threshold that makes the negative prediction value reach 100% as the first threshold, and select the first threshold that makes the positive prediction value reach 100% as the second threshold.

7. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 1, wherein: Diagnostic performance validation included calculating the area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, negative predictive value, calculating the puncture avoidance rate for the low-risk group and the correct identification rate for the high-risk group, and comparing the statistical differences between the integrated model and the single model using the DeLong test.

8. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 4, wherein: It also includes the use of SHAP to analyze the contribution of each feature to the prediction results of the integrated model and visualize the importance of features. Among them, prostate health index density, prostate imaging report and data system score, and prostate volume are the core features that affect the model output.

9. The three-category risk stratification method for prostate cancer based on an ensemble learning model according to claim 3, wherein: LASSO regression achieves variable screening by minimizing the objective function, which is: Among them, β is the regression coefficient to be optimized; n is the number of training samples; y i is the true label of the i-th sample; is the predicted probability of the i-th sample, X is the sample feature matrix; the objective function is solved by the coordinate descent method, the optimal regularization parameter λ is determined by cross-validation, and finally the features with non-zero coefficients are retained as prediction features.

10. A three-category risk stratification system based on an ensemble learning model, characterized in that: The system is used to implement the method according to any one of claims 1 to 9, and the system includes: The data processing module is used to collect clinical information and pathological results of patients with elevated PSA, perform data preprocessing, and screen clinical characteristics through LASSO regression; The model building module is used to build multiple different types of machine learning base models, and to build soft voting ensemble models and stacking ensemble models through ensemble learning strategy combinations; The risk stratification module is used to set dual-threshold cutoff points based on the predicted probability of the integrated model, establish dual-threshold stratification rules, and divide patients into three levels: low risk, medium risk, and high risk; The result output module is used to verify the diagnostic performance of the three-category risk stratification model on the test set and generate a model performance evaluation report.

Images