Open-Source AI Revolutionizes CT Body Composition Analysis for Clinical Use
By Yaqian Chen, Hanxue Gu, Yuwen Chen, Jichen Yang, Haoyu Dong, Joseph Y. Cao, Adrian Camarena, Christopher Mantyh, Roy Colglazier, Maciej A. Mazurowski
Published on November 24, 2025| Vol. 1, Issue No. 1
Content Source
This is a curated briefing. The original article was published on cs.CV updates on arXiv.org.
Summary
This briefing introduces a new, publicly available, end-to-end AI model designed for automated, comprehensive body composition analysis using axial CT images. The model accurately segments skeletal muscle, subcutaneous adipose tissue (SAT), and visceral adipose tissue (VAT) across the chest, abdomen, and pelvis, while also calculating critical metrics such as muscle density, visceral-to-subcutaneous fat (VAT/SAT) ratio, and skeletal muscle index (SMI) for both 2D and 3D assessments. Evaluated on diverse internal and external datasets, the model demonstrates high performance, achieving Dice coefficients exceeding 89% for all segmentations and mean relative absolute errors under 10% for all metrics, significantly outperforming benchmarks and addressing a critical gap in accessible clinical tools.
Why It Matters
This development represents a significant leap for AI in healthcare, particularly in medical imaging. The public release of a high-performing, end-to-end body composition segmentation and feature calculation model democratizes access to advanced diagnostic capabilities previously limited to specialized in-house solutions. For AI professionals, this underscores a powerful trend: the transition of robust research tools into practical, open-source applications that can drive tangible clinical impact. It enables more consistent and scalable assessment of crucial health indicators like cardiovascular risk, metabolic health, nutritional status, and oncology treatment response, moving precision medicine forward. Furthermore, the emphasis on rigorous validation across diverse age, sex, and race groups highlights the growing importance of generalizability and fairness in medical AI models. By making such a tool accessible, it fosters collaboration, accelerates further research and development in quantitative imaging biomarkers, and sets a precedent for how AI can empower clinicians with automated, data-rich insights from standard imaging, ultimately improving patient outcomes and streamlining workflows.