Towards Adaptive Cancer Treatment: A Review of Explainable and Agentic AI in Iterative, Multimodal Healthcare Models

The integration of Artificial Intelligence (AI) into cancer diagnosis and treatment has transformed oncology workflows by enhancing precision, efficiency, and decision support. However, most AI implementations remain static, failing to accommodate the dynamic and complex nature of individual patient trajectories. This review examines the limitations of current AI approaches in oncology and proposes the fusion of Explainable AI (XAI) and Agentic AI to address these challenges. We explore iterative learning, multimodal data integration, ethical agentic behavior, and feedback-driven reasoning to advance personalized oncology care. By reviewing 30 recent AI applications across various cancer types, we identify significant research gaps and propose a novel hybrid framework combining context-aware agentic reasoning with explainable decision-making. Our findings emphasize the necessity for adaptive, transparent, and ethically-aligned AI systems to support real-time clinical decisionmaking and improve cancer outcomes.