Objectives: Artificial Intelligence (AI) has the potential to transform clinical medicine by enhancing diagnostic precision, prognostic accuracy, and personalized decision making. In urological oncology, AI technologies are increasingly integrated across diagnostic, prognostic, and therapeutic pathways. This overview of systematic reviews aims to synthesise the current evidence on the performance and clinical utility of AI applications in urological malignancies. Methods: A systematic search was conducted of major bibliographic databases from 1974 to October 2024 to identify systematic reviews (SRs) evaluating AI-based models in urological cancers. We followed the Cochrane methodology of Overviews of Reviews. Outcomes of interest included diagnostic accuracy metrics (area under the receiver operating characteristic curve AUC-ROC, sensitivity, specificity, and accuracy) and predictive performance for treatment-relatedoutcomes. Results: A total of 67 SRs encompassing 1139 primary studies were included. RoB was assessed as low in 21 reviews, high in 41, and unclear in 5. Forty-five SRs evaluated AI in prostate cancer (PCa). AI models demonstrated high accuracy (88–93%) in distinguishing benign from malignant lesions, and outperformed human interpretation alone in predicting clinically significant PCa (AUC: 0.933 vs. AUC: 0.82–0.87). Performance in Gleason score classification was also very strong (AUC: 0.72–0.99, Accuracy: 70.8–93.0%). Sixteen SRs assessed AI in Renal carcinoma (RCa), reporting accurate subtype classification and tumor grade prediction, with some models exceeding 90% accuracy. Sixteen SRs focused on Bladder Cancer (BCa), showing improved detection of muscle-invasive disease and recurrence risk prediction compared with the conventional approach. Key limitations included heterogeneity in validation strategies, overlap of primary studies across reviews, and potential oversampling due to synthetic or augmented data. Conclusions: AI-based models consistently demonstrate improved diagnostic and predictive performance across urological malignancies and hold promise for enhancing clinical decision-making and patient outcomes. At present, AI tools should be regarded as decision-support systems that complement, rather than replace, clinical expertise. Future research should emphasize prospective validation, standardized outcome definitions, and evaluation of real-world clinical impact. Robust and harmonized evaluation frameworks are also essential to address challenges related to data quality, privacy, generalizability, and ongoing regulatory and legal uncertainty, thereby enabling safe and equitable integration of AI into routine urological cancer care.
Sakalis et al. (Thu,) studied this question.