From Patient Selection to Surveillance: Artificial Intelligence Applications in Radioiodine Therapy – A Systematic Review

Abstract Radioactive iodine therapy requires individualized decisions regarding patient selection, administered activity planning, imaging assessment, and follow-up. This systematic review summarizes current applications of artificial intelligence (AI) in the 131 I treatment pathway and evaluates barriers limiting clinical implementation. Following PRISMA 2020 guidelines, PubMed, Scopus, Web of Science, and Wiley databases were searched for original studies involving human patients treated with 131 I for differentiated thyroid cancer or benign thyroid disease, where AI, deep learning (DL), or machine learning (ML) methods were utilized. Evidence was synthesized, and study quality was assessed using the NIH/NHLBI Study Quality Assessment Tool. Among 775 identified records, 24 studies met inclusion criteria. AI applications included pretreatment decision support, detection and characterization of iodine-avid disease, dosimetry and activity optimization, radiation-safety assessment, treatment response prediction, levothyroxine dose estimation, and recurrence-risk stratification. The available evidence was dominated by retrospective, single-center studies, primarily focused on differentiated thyroid cancer, with limited external validation. AI has the potential to improve standardization and personalization of 131 I therapy, but heterogeneous methodologies and insufficient evidence of clinical utility currently restrict routine adoption. Prospective multicenter studies with harmonized endpoints and robust validation are required to facilitate clinical translation.