Sarcopenia, a debilitating skeletal muscle disorder fueling frailty, disability, and mortality in aging societies, is increasingly targeted by cutting-edge artificial intelligence (AI) for enhanced detection and evaluation. This comprehensive review spotlights AI-driven body shape analysis for sarcopenia, spanning medical imaging techniques (CT, MRI, DXA, ultrasound) and innovative 3D body surface scanning. Drawing from 962 publications across key databases (January 2015 to August 2025), our methodology incorporated search terms for medical imaging alongside 3D body shape, morphometry, and computer graphics. Strikingly, while encompassing these expansive terms, results underscore AI's heavy reliance on medical imaging, leaving 3D scanning as a promising yet untapped avenue for accessible, non-invasive screening. Studies chiefly leverage convolutional neural networks for precise muscle segmentation at L3, machine learning models (Random Forest, SVM, XGBoost) for risk forecasting, and deep learning for skeletal muscle index computation, aligned with standards like EWGSOP2 and AWGS 2019. Highlights include AI's prowess in boosting diagnostic accuracy via automated, consistent metrics and unlocking opportunistic screening from everyday scans. Yet, hurdles persist in data uniformity, inclusivity, model transparency, and real-world testing. Looking ahead, priorities encompass explainable AI for clarity, harnessing 3D body analysis for broader reach, and rigorous prospective studies for seamless clinical adoption. Setting this apart from static reviews, we introduce a dynamic companion website that auto-updates with fresh references daily; its open-source code serves as an adaptable template for other domains via simple keyword and scope configurations, while a fixed compilation of references from this review ensures perpetual currency.
Download Full PDF@article{aiersilan2026literature,
title={Literature Review of AI-Driven Body Shape Analysis for Sarcopenia},
author={Aiersilan, Aizierjiang and Hahn, James},
journal={Authorea Preprints},
year={2026},
publisher={Authorea}
}