Artificial Intelligence in Radiation Therapy, Imaging and Radiation Protection: A Comprehensive Review

Artificial intelligence (AI) has rapidly evolved from experimental prototypes to clinically relevant tools across radiation therapy, diagnostic imaging, and radiation protection. In radiation therapy, deep learning (DL) enables auto-segmentation, dose prediction, adaptive MR-linac workflows, and data-driven quality assurance, demonstrating measurable efficiency gains and improved consistency. In imaging, AI-based reconstruction techniques reduce CT radiation dose by up to 40–45% while preserving diagnostic accuracy, accelerate MRI acquisition without loss of fidelity, and support PET imaging at significantly reduced counts using transformer-based models. In radiation protection, AI-driven pipelines enable personalized organ dosimetry, real-time staff exposure monitoring, and decision-support systems that enhance safety in interventional suites.

Despite these advances, critical challenges persist, including dataset shift, limited prospective trials, workflow integration barriers, and the need for uncertainty quantification. Regulatory frameworks such as the EU AI Act and FDA pathways, along with ethical guidance from WHO, are shaping deployment toward safe, transparent, and accountable use. This comprehensive review synthesizes current evidence, highlights mature and emerging applications, and outlines limitations and future directions to ensure sustainable and equitable adoption of AI in clinical practice.