Clinical Report: Combining Radiomic Features from Biparametric MRI with Clinical Data
Overview
This study evaluates the integration of radiomic features from biparametric MRI with clinical variables to predict biochemical recurrence (BCR) after radical prostatectomy. The combined model demonstrated performance compared to traditional clinical models.
Background
Prostate cancer is the most commonly diagnosed cancer in men, with a significant proportion experiencing biochemical recurrence after radical prostatectomy. Current clinical risk stratification tools often fail to capture the biological complexity of tumors.
Data Highlights
Model
AUC
Calibration
Net Benefit
Combined Model
0.85 (95% CI 0.83–0.87)
Slope = 1.01; Brier = 0.13
Higher than D’Amico classification
Radiomics-only
0.78
N/A
N/A
Clinical-only
0.72
N/A
N/A
Key Findings
The combined model achieved an AUC of 0.85, outperforming radiomics-only (0.78) and clinical-only (0.72) models.
Calibration of the combined model was strong with a slope of 1.01 and a Brier score of 0.13.
Decision-curve analysis indicated a higher net benefit for the combined model compared to the D’Amico classification.
The most influential predictors included Gleason Grade Group, PSA, and radiomic first-order/texture features.
Clinical Implications
The integration of radiomic features from bpMRI with clinical variables may enhance pre-treatment predictions for biochemical recurrence in prostate cancer patients.
Conclusion
Further validation in independent cohorts is required before clinical implementation.
by Selma Bozorgpana, Indri Desiati, Mohammed R. S. Sunoqrot, Petter Davik, Guro F. Giskeødegård, Gabriel Addio Nketiah, Mattijs Elschot, May-Britt Tessem, Tone F. Bathen
