Artificial intelligence in prostate cancer screening - a narrative review of diagnostic and clinical implications

Abstract

Background. Current Prostate cancer screening methods have major drawbacks, such as high rates of false-positive (70%), failure to detect clinically relevant tumors (20-30%), and considerable inter-observer variability. Artificial intelligence holds promise to enhance outcomes in the entire screening process. Objective(s). To critically examine AI integration in prostate cancer screening, analyzing its diagnostic performance, clinical relevance, and implementation challenges to guide informed, evidence-based adoption. Materials and methods. A narrative review using a systematic search of major databases (PubMed/MEDLINE, Scopus, Google Scholar) following the SANRA guidelines. Studies focusing on AI applications in PSA analysis, MRI interpretation, Histopathology, and multimodal diagnostic methods from 2018 - 2025 were included. Study quality assessment used adapted QUADAS-2 criteria. Results. AI outperformed traditional approaches across all diagnostic modalities. PSA Analysis (0.82-0.89 vs 0.59-0.63 for conventional methods), digital pathology (97.4% sensitivity,94.8% specificity), and MRI methods (AUC 0.91 vs 0.86 for radiologists). The FDA-approved Paige system enhanced pathologist sensitivity by 8% while preserving specificity. In the PI-CAI trials (n=10207 MRIs), AI diagnosed 6.8% more clinically relevant tumors at equivalent specificity. Combined AI systems improved risk stratification by over 13-15%. Economic analysis also showed cost-effectiveness and its potential to reduce diagnostic workload by 60-80%. Conclusion(s). AI is clinically ready for prostate cancer screening, supported by extensive validation studies. Effective implementation depends on infrastructure, workflow integration, and clinician training. AI can enhance diagnostic accuracy and reduce unwanted procedures, advancing prostate cancer screening.