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

Abstract

Background: To establish a machine learning (ML) model for predicting prostate biopsy outcomes using prostate-specific antigen (PSA) values, multiparametric magnetic resonance imaging (mpMRI) findings, and hematologic parameters. Methods: The medical records of the patients who had undergone a prostate biopsy were evaluated. Laboratory findings, mpMRI findings, and prostate biopsy results were collected. Patients with benign prostate pathology were classified as Group 1, and those with prostate cancer (PCa) were classified as Group 2. The following ML algorithms were used to create the ML model: ExtraTrees classifier, Light Gradient-Boosting Machine (LGBM) classifier, eXtreme Gradient Boosting (XGB) classifier, Logistic Regression, and Random Forest classifier. Results: A total of 244 male patients who met the inclusion criteria were included in this study. Among them, 171 (71.1%) were categorized in Group 1, and 73 (29.9%) in Group 2. The LGBM classifier model demonstrated the highest performance, achieving an accuracy rate of 81.6% and an AUC–ROC (area under the curve–receiver operating characteristic) of 78.4%, with sensitivity and specificity values of 66.7% and 88.2%, respectively, in predicting prostate biopsy outcomes. Conclusions: Pathological results can be predicted by ML models using PSA values, mpMRI findings, and hematologic parameters prior to a prostate biopsy, potentially reducing unnecessary biopsy procedures.