Next-Generation Foundation AI Models for Precision Radiomics: Unlocking High-Dimensional Imaging Biomarkers for Precision Oncology

Abstract

Next-generation foundation artificial intelligence (AI) models are redefining precision radiomics by enabling comprehensive extraction, integration, and interpretation of high-dimensional imaging biomarkers for personalized cancer diagnosis, prognostic prediction, therapeutic optimization, and longitudinal disease monitoring. Conventional radiomics has demonstrated considerable potential for quantitative characterization of tumor heterogeneity; however, handcrafted imaging features, limited generalizability, fragmented multimodal integration, and inadequate biological interpretability have constrained widespread clinical implementation. Recent advances in foundation AI models, multimodal transformer architectures, graph neural networks, self-supervised learning, reinforcement learning, and generative artificial intelligence have enabled automated learning of generalized imaging representations from computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), ultrasound, mammography, digital pathology, genomics, transcriptomics, proteomics, metabolomics, epigenomics, spatial biology, laboratory biomarkers, circulating tumor DNA, electronic health records, biomedical literature, and longitudinal clinical outcomes. These intelligent systems support radiogenomic biomarker discovery, molecular characterization, prognostic prediction, treatment response assessment, immunotherapy selection, digital twin simulation, adaptive disease monitoring, and evidence-based clinical decision support. Emerging technologies including multimodal large language models, agentic AI, federated learning, explainable artificial intelligence, retrieval-augmented generation, cancer knowledge graphs, cloud-native healthcare platforms, and digital health ecosystems further strengthen precision radiomics by enabling collaborative, privacy-preserving, transparent, and continuously adaptive biomedical intelligence. Despite remarkable technological advances, important scientific, technical, ethical, and regulatory challenges remain regarding imaging standardization, multimodal harmonization, computational scalability, interoperability, cybersecurity, clinical validation, and equitable implementation. This review provides a comprehensive overview of next-generation foundation AI models for precision radiomics, emphasizing their role in unlocking high-dimensional imaging biomarkers for precision oncology.[1]