Related Experiment Video
Updated: Sep 13, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
378
A visualized machine learning model using noninvasive parameters to differentiate men with and without prostatic
Wenting Zhou1, Linhui Wang2, Xue Zhang1
1Department of Pathology, The First People's Hospital of Longquanyi District, Chengdu, 610100, China.
Scientific Reports
|July 27, 2025
Summary
A new visualized XGBoost model accurately distinguishes prostate cancer (PCA) from benign conditions using noninvasive parameters. This AI tool aids in selecting patients for biopsy, potentially reducing unnecessary procedures.
Area of Science:
- Urology
- Oncology
- Artificial Intelligence in Medicine
- Machine Learning
Background:
- Accurate differentiation between prostate carcinoma (PCA) and benign prostatic hyperplasia (BPH) is crucial for appropriate patient management.
- Noninvasive prebiopsy parameters offer a potential avenue for improving diagnostic accuracy and reducing the need for invasive procedures.
- Machine learning models show promise in analyzing complex datasets for medical diagnosis.
Purpose of the Study:
- To develop and validate a visualized extreme gradient boosting (XGBoost) model for distinguishing PCA from non-PCA using noninvasive prebiopsy parameters.
- To compare the performance of the XGBoost model against other machine learning and logistic regression models.
- To assess the potential of the model in aiding clinical decision-making for prostate biopsy selection.
Main Methods:
- A cross-sectional study involving 310 Chinese men undergoing prostate biopsy.
- Analysis of 15 noninvasive prebiopsy parameters using the XGBoost algorithm.
- Model performance evaluated using the area under the receiver operating characteristic curve (AUC) and compared with decision tree, lasso, neural network, support vector machine, and logistic models.
Main Results:
- The visualized XGBoost model achieved a high AUC of 0.965, significantly outperforming other models (AUC range: 0.708–0.817) and the logistic model (AUC: 0.813).
- Key predictors identified included serum thymidine kinase 1 (STK1p), total prostate-specific antigen (TPSA), age, free prostate-specific antigen (FPSA), and free-to-total prostate-specific antigen (FTPSA).
- The model utilized eight noninvasive predictors and generated 49 visualized decision trees to aid interpretation.
Conclusions:
- A visualized XGBoost model demonstrates high accuracy in differentiating PCA from non-PCA using noninvasive prebiopsy parameters.
- This AI-driven approach can assist in the precise selection of high-risk PCA patients for biopsy.
- The model has the potential to minimize unnecessary biopsies, associated costs, and patient discomfort.
