Researchers are exploring non-contrast MRI, or bpMRI, as a tool for clinically significant prostate cancer screening. Furthermore, artificial intelligence (AI) is emerging as a critical support for radiologists who have less experience. Consequently, a new study within the PROSA trial evaluated how effectively AI software assists these professionals. The team analyzed 499 bpMRI cases using PI-RADS v2.1-compliant protocols. Experts compared the performance of less-experienced readers with and without assistance from AI Prostate MRI screening tools. Additionally, they compared these results against the AI software working alone. Therefore, the researchers could clearly see the benefits of human-AI collaboration.
The AI-assisted reader achieved the highest diagnostic performance overall. Specifically, this combination reached a sensitivity of 76.5% and an accuracy of 95.8%. In contrast, the unaided reader only achieved 93.2% accuracy. Moreover, the software reduced the number of equivocal PI-RADS 3 cases significantly.
Advancing Accuracy with AI Prostate MRI Screening
Integrating AI into the workflow improves inter-reader agreement. For example, the agreement score jumped from 0.64 to 0.84 with software help. This improvement ensures that different doctors reach similar conclusions. Furthermore, the exact agreement with experts rose from 32.5% to 54.5% when using AI. The researchers concluded that AI enhances consistency in bpMRI interpretation. It helps prioritize suspicious cases in high-volume settings. Consequently, this technology can alleviate the reporting burden on busy departments. Finally, it supports equitable early detection by providing expert-level insights to all readers.
Frequently Asked Questions
Q1: How does AI assist in interpreting prostate MRI?
AI provides decision support by highlighting suspicious lesions. This reduces the number of equivocal PI-RADS 3 cases and improves accuracy for less experienced readers.
Q2: What were the main performance metrics in the study?
The AI-assisted reader achieved 76.5% sensitivity and 95.8% accuracy. This combination outperformed both unaided human readers and the AI software working alone.
References
- Messina E et al. Artificial intelligence-assisted reading of non-contrast prostate MRI: application and concordance with expert interpretation in a screening population within the PROSA trial. Eur Radiol. 2026 Feb 13. doi: 10.1007/s00330-025-12269-7. PMID: 41686238.
- Drzadeclinic.com. Benefits of AI for Prostate Cancer Detection on Multiparametric MRI. Published May 24, 2023.
- MedicalDialogues.in. Risk Stratification and Artificial Intelligence in Early Magnetic Resonance Imaging-based Detection of Prostate Cancer. Published January 5, 2022.