In oncology, body composition (BC) provides clinically meaningful information beyond body mass index, capturing muscle and adipose tissue alterations associated with survival, treatment tolerance, surgical complications and quality of life. Although routine oncologic imaging is widely available, BC assessment remains poorly integrated into daily clinical practice, largely because conventional imaging-based approaches require time-consuming manual analyses, dedicated software and specialized expertise. Artificial intelligence (AI), particularly deep learning-based image segmentation, may automate BC analysis and generate rapid, reproducible, and scalable estimates from routinely acquired imaging, without increasing clinical workload. This opinion paper aims to examine AI-based BC analysis as a potential strategy to integrate BC into routine oncology workflows, outlining its potential clinical benefits and the aspects that need to be addressed before widespread implementation. AI-based BC analysis may improve nutritional assessment, refine clinical and nutritional risk stratification, and help identify patients at increased risk of treatment-related toxicity. In perspective, BC data may also support more personalized nutritional and physical activity interventions and contribute to muscle mass-informed anticancer treatment dosing strategies. Several gaps still limit its clinical implementation, including the need of robust external validation, standardized acquisition and analytical protocols, clinically meaningful cut-offs and ethical, and regulatory and data governance frameworks. AI-based BC analysis is therefore a promising but still evolving approach that may help translate BC from a prognostic marker into a clinically actionable tool in oncology.
Mattavelli et al. (2026) studied this question.