This thesis examines the development and validation of AI systems for prostate cancer detection on MRI, with the aim of generating robust evidence to guide their integration into clinical workflows. To this end, it examined three central questions: (1) whether deep learning-based systems can autonomously localize and classify prostate cancer in 3D; (2) whether state-of-the-art AI systems can achieve non-inferior diagnostic performance to international radiologists and the standard of care; (3) whether such systems can be deployed safely and reliably across global, multiethnic populations and diverse healthcare settings.

Go to thesis