mpMRI-Based Risk Estimation to Optimize Prostate Cancer Patient Selection for Active Surveillance.

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Active surveillance (AS) is a viable option for low-risk prostate cancer (PCa).
• Current risk stratification tools are insufficient for optimal AS patient selection.
• Early detection of PCa progression is crucial for effective management.

Purpose of the Study:

• To develop an integrated method for predicting PCa progression within 12 months.
• To enhance patient selection for AS by categorizing individuals into rapid and slow progressors.
• To improve the safety and efficacy of AS strategies.

Main Methods:

• Utilized convolutional neural networks for multiparametric MRI (mpMRI) lesion segmentation.
• Integrated mpMRI radiomics with clinical variables (age, PSA, PI-RADS) for risk prediction.
• Trained models on radical prostatectomy data mapped to mpMRI and prospectively validated on 163 participants.

Main Results:

• The clinical-radiomics model achieved an AUC of 0.84 in predicting progression.
• The model significantly improved AS patient selection in an independent test set.
• Negative predictive value increased by 18.5% compared to standard care (p < 0.001).

Conclusions:

• The risk assessment platform reliably differentiates AS candidates with stable disease from those likely to progress early.
• This tool shows promise for use during annual follow-up visits.
• Improved risk stratification can optimize AS management for prostate cancer patients.