A NEW artificial intelligence–driven MRI system has shown strong potential to improve the detection of clinically significant prostate cancer while reducing radiology workload. The automated tool, known as ProAI, was developed to address longstanding challenges associated with variability in PI-RADS scoring, a widely used but subjective system for interpreting prostate MRI scans.
The research team trained, validated, and tested ProAI using 7,849 MRI examinations collected from six centres and two public imaging datasets. The system analyses biparametric MRI, an increasingly adopted, shorter, and contrast-free imaging protocol, and generates a patient-level risk estimate for clinically significant prostate cancer (csPCa).
AI-MRI Decision Aid for Prostate Cancer Detection
Across pooled external test datasets, the tool achieved an area under the receiver operating characteristic curve (AUC) of 0.93 (95% CI 0.91–0.95), demonstrating diagnostic performance comparable to PI-RADS while improving consistency between cases. Variability in interpretation is a known limitation of traditional scoring, often impacting reproducibility across institutions and experience levels.
In a multi-reader, multi-case study involving nine clinicians, ProAI further improved diagnostic accuracy, increasing reader performance from 0.80 to 0.86 when used as a decision aid. Importantly, its use significantly reduced time spent interpreting scans, suggesting meaningful workflow advantages.
Real-world performance was assessed through prospective implementation in 1,978 consecutive MRI examinations. ProAI maintained high diagnostic accuracy (AUC 0.92) and contributed to a 32% reduction in radiology workload, an important consideration amid rising demand for prostate imaging and workforce shortages. The system also generalised well to external public datasets, including the TCIA cohort, where it achieved an AUC of 0.83.
Implications for Prostate Cancer Care Pathways
The authors suggest that automated MRI decision aids like ProAI can support standardised reporting, enhance diagnostic efficiency, and streamline cancer-care pathways. These findings offer promising evidence that AI can help address key bottlenecks in prostate cancer imaging.
Prospective evaluation was conducted as part of a registered clinical trial (ChiCTR2400092863), further supporting the robustness of the study.
Reference
Wu H et al. Automated MRI system for clinically significant prostate cancer detection development validation and real-world implementation. 2025. Nat Comm. 2025; doi:10.1038/s41467-025-66593-z
