Beyond the field of view: Radiomics and artificial intelligence for organ-specific and systemic disease prediction from localized medical imaging.
Authors
Affiliations (5)
Affiliations (5)
- Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, LA, CA, 90048, USA.
- Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, LA, CA, 90048, USA; Department of Bioengineering, University of California, Los Angeles, CA, 90048, USA.
- Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, CA, 90048, USA.
- Division of Digestive and Liver Diseases, Cedars-Sinai Medical Center, Los Angeles, CA, 90048, USA.
- Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, LA, CA, 90048, USA. Electronic address: [email protected].
Abstract
Radiomics and artificial intelligence (AI) are increasingly used in medical imaging to shift beyond qualitative interpretation toward quantitative disease prediction. Localized medical imaging, including abdominal computed tomography (CT), chest CT, mammography, brain magnetic resonance imaging (MRI), and retinal imaging, may reveal subtle features that reflect not only pathology within the imaged organ but also systemic or distant disease processes. By extracting radiomic features with machine learning or learning imaging representations with deep learning models, previous studies have shown predictive value across neurology, oncology, cardiometabolic disease, and biological aging. This narrative review summarizes recent advances in radiomics and AI for organ-specific and systemic disease prediction from localized imaging. It outlines key methodological steps, including image acquisition, segmentation, harmonization, feature extraction, model development, validation and clinical translation. This review also describes the biological basis linking localized imaging biomarkers to systemic disease processes, such as metabolic dysfunction, inflammation, vascular remodeling, and biological aging. Ongoing challenges include imaging variability, limited generalizability, confounding, reproducibility, interpretability, ethical considerations, and the need for prospective validation. Overall, imaging-based biomarkers have the potential to extend the value of routine scans by enabling earlier risk detection, more individualized risk stratification and targeted follow-up beyond the original field of view.