Artificial intelligence and radiomics in bladder cancer MRI: a scoping review of applications, performance, and barriers to clinical translation

Comprehensive Summary

Bladder cancer diagnosis and treatment rely on accurate determination of tumor stage and grade using multiparametric magnetic resonance imaging (mpMRI), standardized by VI-RADS. This review evaluates the potential of AI in assisting in this process, by reducing variability in human interpretation. While reported performance is often very high, with an AUC up to 0.99, more reliable multi-center studies suggest a more realistic AUC of 0.85-0.95, positioning AI as a compliment to VI-RADS rather than a replacement for radiologists. A major limitation highlighted is the gap between technical performance and clinical application, as most studies focus on diagnostic accuracy rather than whether AI is improving patient outcomes. Overall, despite promising results, more multi-center studies that focus on real clinical impact are needed before integrating AI into healthcare.

Outcomes and Implications

AI has strong potential to improve bladder cancer diagnosis and treatment through imaging, but without more reliable studies with evidence of clinical benefit, it cannot be reliably integrated into healthcare.