Multimodal Artificial Intelligence in Oncology: Integrating Radiomics, Pathomics, and Genomics

Abstract

The emergence of artificial intelligence (AI) has fundamentally transformed modern oncology by enabling the computational interpretation of increasingly complex biomedical datasets derived from radiology, digital pathology, genomics, transcriptomics, and electronic health records. The integration of these heterogeneous data sources has facilitated the development of multimodal AI frameworks capable of generating biologically meaningful and clinically actionable insights across the entire cancer care continuum. Unlike unimodal computational approaches that analyze isolated data modalities, multimodal AI combines radiomics, pathomics, and genomics to comprehensively characterize tumor heterogeneity at anatomical, histopathological, molecular, and temporal levels. Recent advances in deep learning, transformer architectures, self-supervised learning, and foundation models have substantially enhanced the capacity of multimodal systems to identify latent biological patterns associated with tumor initiation, disease progression, therapeutic resistance, and patient survival. These technologies are increasingly being applied to early cancer detection, molecular subtyping, prognostic risk stratification, immunotherapy response prediction, and precision therapeutic decision-making. Despite these advances, several translational challenges remain, including inadequate data harmonization, algorithmic bias, limited model interpretability, regulatory uncertainty, and the lack of large-scale prospective clinical validation. In addition, ethical considerations surrounding data governance, privacy preservation, and equitable implementation continue to influence the clinical adoption of AI-driven oncology solutions. This review critically examines the evolution of multimodal AI in oncology, with particular emphasis on radiomics, pathomics, and genomics integration strategies, transformer-based architectures, explainable AI, and multimodal fusion techniques. The article further explores contemporary clinical applications, translational challenges, and emerging opportunities involving foundation models and federated learning within precision oncology. Collectively, multimodal AI represents a transformative paradigm with the potential to redefine oncologic diagnosis, personalize therapeutic interventions, strengthen clinical decision support, and accelerate the advancement of precision cancer medicine.