Making Body Composition Analysis Available in Clinical Routine

Abstract Body composition metrics are essential for assessing patient condition, particularly in the context of sarcopenia, but are often underutilized in clinical practice due to the complexity of image-based analysis. This work presents a modular, Docker-based framework for the automated extraction of sarcopenia-relevant parameters from single-slice CT images at the L3 level. A deep learning model performs multilabel segmentation of key muscle and adipose tissues, including intramuscular adipose tissue (IMAT), enabling the calculation of clinically relevant metrics such as cross-sectional area and tissue density. The system supports both conventional and photon-counting CT data and includes a streamlined graphical user interface designed for ease of use in clinical environments. Results are accessible via a lightweight interface and can be exported in standardized formats for further analysis. The framework is openly available on GitHub and provides a reproducible, secure, and adaptable solution for integrating AI-driven body composition analysis into clinical workflows.