Body composition analysis on routine CT imaging: a predictive health opportunity

Routine computed tomography (CT) examinations contain quantitative information on skeletal muscle, adipose tissue distribution, and bone integrity that is strongly associated with clinical outcomes across oncology, cardiovascular disease, surgery, and critical care. This review aims to synthesize recent advances in artificial intelligence (AI)–enabled body composition analysis on routine CT and to outline its translational pathway toward possible clinical implementation as an opportunistic, value-added imaging biomarker. A narrative review of the literature was performed focusing on CT-based body composition analysis, deep learning–driven segmentation techniques, and their clinical applications. We examined technical foundations of AI segmentation, multidimensional body composition biomarkers (muscle, adipose tissue, ectopic fat, and bone), validation strategies, and real-world workflow integration within radiology reporting systems and electronic health records. Advances in deep learning have enabled fast, reproducible, and scalable segmentation of skeletal muscle, visceral and subcutaneous adipose tissue, ectopic fat depots, and vertebral bone from routine CT scans without additional imaging or radiation. Multidimensional and longitudinal body composition phenotypes further improve risk stratification. Emerging workflows support structured reporting, automated risk flags, and integration with clinical decision support systems, although challenges remain in standardization, external validation, harmonization across scanners and protocols, and demonstration of clinical utility. AI-enabled body composition analysis has the potential to transform routine CT into a platform for physiologic phenotyping and opportunistic risk assessment. Successful clinical adoption will require multicenter validation, standardized reporting frameworks, workflow integration, and alignment with regulatory and reimbursement structures.