Prostate cancer diagnostics in transition From clinical strategies to artificial intelligence

This thesis examines how prostate cancer diagnostics can be optimized by refining biopsy strategies and exploring new imaging technologies. A central finding is that contralateral systematic biopsies in men with a unilateral MRI lesion provide limited additional diagnostic value. They rarely influence surgical or radiotherapy planning and have minimal impact on risk classification. In contrast, combining MRI-targeted biopsies with ipsilateral systematic biopsies improves detection of clinically significant, higher-grade disease, supporting a more focused and less invasive biopsy approach.
The thesis also evaluates MRI quality using the PI-QUAL scoring system. While PI-QUAL can identify clearly inadequate scans, its reproducibility between readers is limited, and its relationship with diagnostic accuracy is less clear in daily practice. This suggests its main role is in quality assurance rather than direct clinical decision-making.
In addition, this work develops and validates an alternative diagnostic pathway based on three-dimensional multiparametric ultrasound (3D mpUS) combined with artificial intelligence. The study demonstrates that automated prostate segmentation and AI-based lesion detection are feasible and accurate. Simulated and prospective analyses show that AI-assisted mpUS achieves detection rates for clinically significant prostate cancer comparable to MRI-based pathways, while detecting fewer low-grade tumors.
However, the main limitation of mpUS is not accuracy but feasibility, as technical and quality issues currently restrict its routine use. Overall, this thesis shows that prostate cancer diagnostics can be streamlined and diversified, reducing unnecessary procedures while maintaining diagnostic performance.