Adaptive Radiotherapy in the Era of Image Guidance and Artificial Intelligence: A Critical Review

Adaptive radiotherapy (ART) transforms radiotherapy from a static treatment plan into a monitored, feedback-driven process that can respond to anatomic and biologic change. This critical narrative review synthesizes foundational physics literature, consensus reports, prospective studies, randomized trials, implementation reports, and ART-specific artificial-intelligence validation studies, with emphasis on image guidance, evidence quality, and clinical actionability. Further the review separates image guidance, reduced margins, gating, tracking, plan-of-the-day selection, offline replanning, and online reoptimization because these related interventions have different evidence bases. ART is most compelling when anatomic change is frequent, dosimetrically important, measurable, and not adequately managed by population margins or conventional image guidance. Quantitative evidence is strongest for dosimetric recovery, organ-at-risk sparing, workflow feasibility, and selected toxicity reduction; evidence for survival improvement remains limited and heterogeneous. Offline ART has mature roles for progressive or systematic change, whereas online MR-guided and cone-beam CT-guided workflows address daily anatomy through same-session contour review, plan selection or reoptimization, and rapid quality assurance. Artificial intelligence can accelerate segmentation, synthetic CT generation, planning, and quality assurance, but clinically meaningful validation requires geometric, dosimetric, time, failure-rate, bias, and post-update monitoring. ART should be deployed through explicit adaptation triggers, validated dose-accumulation methods, independent quality management, and prospective endpoints connecting adaptive decisions to toxicity, tumor control, workflow efficiency, cost, access, and equity.