Integrating Radiomics and Radiogenomics to Advance Precision Cancer Diagnostics and Treatment

Abstract

Radiogenomics has emerged as a transformative interdisciplinary field that integrates quantitative imaging, genomics, artificial intelligence (AI), and multi-omics data to advance precision oncology. The extraction of quantitative imaging biomarkers from modalities such as computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), magnetic resonance spectroscopy (MRS), and digital pathology assists in the characterization of tumor phenotype, intratumoral heterogeneity, and microenvironmental alterations beyond conventional visual interpretations, which is enabled by radiomics. Integrating the genomic, transcriptomic, proteomic, metabolomic, and clinical data-derived features with imaging provides a non-invasive framework for understanding molecular alterations, disease progression, treatment response, and recurrence patterns. Radiogenomics analysis improved significantly with the artificial intelligence approaches, which include machine learning, deep learning, transfer learning, graph neural networks, and explainable AI by enabling automated feature extraction, multimodal data integration, and predictive modeling. They have also demonstrated substantial clinical relevance in glioblastoma, non-small cell lung cancer, breast cancer, prostate cancer, liver cancer, and brain metastases. Furthermore, the combination of radiomics with the liquid biopsy technologies like circulating tumor DNA and circulating tumor cells, has increased the monitoring of the non-invasive tumor and the dynamic assessment of treatment response. The AI-driven clinical decision support system that is built upon the radiogenomics frameworks is enhancing and supporting risk stratification, adaptive therapy planning, and personalized therapeutic interventions. There are significant barriers, even though there is considerable progress, that continue to limit large-scale clinical implementation, including data heterogeneity, lack of standardization, annotation variability, limited external validation, interpretability concerns, privacy preservation, and evolving regulatory requirements. The developing fields like federated learning, foundation models, digital twins, and real-time AI-enabled imaging workflows are expected to improve these challenges and the scalability, robustness, and clinical translation of radiogenomic systems. Together, the integration of radiomics, radiogenomics, and AIdriven image-omics by enabling comprehensive, non-invasive, and data-driven approaches for cancer diagnosis, prognostication, treatment planning, and longitudinal disease monitoring redefines precision oncology.