Foundation Models for Precision Oncology: Integrating Multimodal Cancer Intelligence Across the Clinical Workflow

Abstract

Cancer remains one of the leading causes of morbidity and mortality worldwide despite remarkable advances in molecular diagnostics, targeted therapeutics, immunotherapy, and precision medicine. The exponential growth of biomedical data generated from radiological imaging, digital pathology, genomic sequencing, transcriptomics, proteomics, metabolomics, laboratory investigations, wearable technologies, and electronic health records has created unprecedented opportunities for artificial intelligence (AI)-driven precision oncology while simultaneously presenting major computational challenges. Foundation models have emerged as a transformative paradigm capable of learning generalized biomedical representations from large-scale multimodal datasets through self-supervised pretraining. Unlike conventional machine learning systems designed for specific clinical tasks, foundation models acquire transferable knowledge that can be adapted across cancer diagnosis, prognostic prediction, molecular characterization, therapeutic optimization, radiogenomics, computational pathology, digital twins, clinical decision support, and personalized medicine. Advances in transformer architectures, multimodal learning, contrastive learning, graph neural networks, large language models, retrieval-augmented generation, agentic AI, and generative artificial intelligence have accelerated development of intelligent oncology foundation models capable of integrating imaging, pathology, multi-omics, longitudinal clinical information, and real-world healthcare data into unified patient-centered computational representations. Despite substantial progress, important challenges remain regarding multimodal data harmonization, computational scalability, explainability, interoperability, cybersecurity, regulatory validation, and equitable implementation. This review provides a comprehensive overview of foundation models in precision oncology, emphasizing computational principles, current clinical applications, implementation challenges, and future perspectives for integrating multimodal cancer intelligence throughout the clinical workflow.[1]