The emergence of radiomics and deep learning technologies has catalyzed a paradigm shift in medical imaging, transitioning from conventional visual assessment to quantitative feature-based mining. Focusing on colorectal cancer, this article systematically delineates the groundbreaking advances in imaging-based artificial intelligence(AI) in supporting tumor diagnosis and treatment decision-making, and critically examines the core challenges encountered during its clinical translation. Looking forward, conducting prospective, multicenter randomized controlled trials to validate the efficacy and safety of imaging AI tools in real-world clinical settings, alongside actively exploring the integration of autonomous intelligent systems into complex clinical workflows, will constitute pivotal pathways for facilitating the clinical deployment of medical imaging AI.