Machine Learning-Based Prediction of Prostate Biopsy Necessity Using PSA, MRI, and Hematologic Parameters

Area of Science:

• Urology
• Medical Informatics
• Machine Learning

Background:

• Prostate cancer (PCa) diagnosis often relies on prostate biopsy.
• Predicting biopsy outcomes non-invasively can improve patient management and reduce healthcare costs.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting prostate biopsy outcomes.
• To integrate prostate-specific antigen (PSA) values, multiparametric magnetic resonance imaging (mpMRI) findings, and hematologic parameters into predictive models.

Main Methods:

• Retrospective analysis of 244 patients who underwent prostate biopsy.
• Collected data included laboratory findings, mpMRI results, and biopsy outcomes.
• Compared ExtraTrees, Light Gradient-Boosting Machine (LGBM), eXtreme Gradient Boosting (XGB), Logistic Regression, and Random Forest classifiers.

Main Results:

• The LGBM classifier achieved the highest performance.
• LGBM model accuracy was 81.6% with an AUC-ROC of 78.4%.
• Sensitivity and specificity for predicting prostate cancer were 66.7% and 88.2%, respectively.

Conclusions:

• Machine learning models can effectively predict prostate biopsy outcomes.
• Pre-biopsy prediction using PSA, mpMRI, and hematologic data shows promise.
• This predictive capability may help reduce the number of unnecessary prostate biopsies.