Artificial intelligence-assisted design and optimization of radiation shielding materials for medical applications

Abstract

The increasing use of ionizing radiation in diagnostic imaging, radiotherapy, nuclear medicine, and interventional procedures has intensified the need for lightweight, effective, and clinically reliable shielding materials. This systematic and critical review evaluates the role of artificial intelligence (AI) in the design, prediction, optimization, and validation of radiation-shielding materials for medical applications. Following the PRISMA 2020 framework, 149 publications published between January 2016 and March 2026 were included from Scopus, Web of Science, IEEE Xplore, ScienceDirect, SpringerLink, and Google Scholar. The evidence was synthesized according to material type, radiation field, dataset characteristics, input descriptors, target properties, AI technique, validation strategy, uncertainty analysis, and experimental maturity. Artificial neural networks, support vector regression, random forests, gradient-boosting methods, and evolutionary algorithms were most frequently applied to attenuation prediction, composition optimization, virtual screening, and Monte Carlo surrogate modeling. The strongest evidence supports interpolation of photon-shielding parameters within defined datasets, whereas external validation, uncertainty quantification, reproducibility, and prospective experimental confirmation remain limited. Many studies also insufficiently distinguish property prediction from genuine materials discovery. AI surrogates can accelerate radiation-transport calculations but should complement rather than replace validated Monte Carlo methods. Clinical translation requires standardized attenuation testing, durability assessment, manufacturing control, toxicological evaluation, and regulatory validation. Future progress depends on representative datasets, grouped and external validation, physics-informed modeling, explainability, and experimentally verified AI-guided material development.