Objective:

To evaluate whether integrating radiomic features from pre-operative biparametric MRI with standard clinical variables improves pre-treatment prediction of biochemical recurrence after radical prostatectomy.

Approach:

• Model Development: A stacked ensemble model was developed using Random Forest and regularized Logistic Regression as base models, with Logistic Regression as the meta-model. The model's performance was evaluated using five-fold stratified cross-validation, SMOTE balancing, Optuna hyperparameter tuning, and isotonic regression-based probability calibration.

Key Findings:

• The combined model achieved an AUC of 0.85 (95% CI 0.83–0.87), outperforming radiomics-only (0.78) and clinical-only (0.72) models.
• Calibration was strong with a slope of 1.01 and Brier score of 0.13.
• High-risk patients had significantly shorter recurrence-free survival (log-rank p<0.001; HR = 5.03).
• The most influential predictors included Gleason Grade Group and PSA, along with radiomic features.

Interpretation:

Integrating bpMRI-derived radiomic features with standard clinical variables improved prediction of biochemical recurrence, providing better discrimination between high-risk and low-risk groups.

Limitations:

• The study is retrospective and conducted at a single center.
• External validation in independent cohorts is required before clinical implementation.

Conclusion:

The results support the role of radiomics in refining individualized recurrence risk assessment in prostate cancer.