Generative Artificial Intelligence for Clinical Decision Support in Oncology: Current Applications, Challenges, and Future Directions

Abstract

Cancer remains one of the leading causes of morbidity and mortality worldwide, creating an urgent need for intelligent clinical decision-support systems capable of delivering personalized, evidence-based cancer care. The rapid expansion of clinical, imaging, molecular, genomic, and radiomics datasets has accelerated the development of generative artificial intelligence (GenAI), enabling advanced computational models to support complex clinical decision-making throughout the oncology care continuum. Powered by transformer-based architectures and large language models (LLMs), GenAI has emerged as a transformative technology with the potential to enhance diagnostic accuracy, optimize treatment selection, facilitate precision oncology, and accelerate translational cancer research. This review provides a comprehensive overview of recent advances in generative artificial intelligence for clinical decision support in oncology, with particular emphasis on large language models, multimodal foundation models, retrieval-augmented generation, and clinical natural language processing. We summarize the underlying computational architectures and examine their applications across cancer screening, diagnosis, radiology, pathology, treatment planning, precision oncology, clinical trial matching, prognostic modeling, patient communication, and multidisciplinary decision-making. Current evidence indicates that GenAI can improve workflow efficiency, automate clinical documentation, synthesize large volumes of biomedical information, support personalized therapeutic recommendations, and facilitate integration of multimodal clinical data into real-time decision-support systems. However, several important barriers remain before widespread clinical implementation can be achieved, including hallucination, limited generalizability across healthcare settings, algorithmic bias, insufficient explainability, privacy and cybersecurity concerns, regulatory uncertainty, and the need for rigorous prospective clinical validation. Future developments are expected to focus on multimodal artificial intelligence, human–AI collaboration, federated learning, foundation models, explainable AI, and continuously learning clinical decision-support systems capable of integrating imaging, pathology, genomics, electronic health records, and real-world patient data. The convergence of generative artificial intelligence and precision oncology has the potential to fundamentally transform clinical decision-making, improve treatment personalization, enhance healthcare efficiency, and ultimately improve outcomes for patients with cancer.