Answers are grounded in this paper and cite their sources.
This review focuses on modern approaches to diagnosing sarcopenia using computed tomography (CT) and machine learning (ML) methods. Sarcopenia, a progressive decline in muscle mass and function, is recognized as a serious complication of many chronic diseases that increases the risk of complications and mortality. Traditional diagnosis is limited by the subjectivity and labor-intensiveness of manual measurements. CT, performed for various clinical indications, has become a key tool for the objective assessment of both the quantity (muscle area, skeletal muscle index – SMI) and quality (density in Hounsfield units) of muscle tissue, predominantly at the level of the third lumbar vertebra (L3). To automate analysis and improve accuracy, ML models, specifically radiomics and deep learning methods, are being actively developed. Radiomics allows for the extraction of hundreds of quantitative features from images reflecting muscle texture and heterogeneity, while convolutional neural networks solve the tasks of automatic muscle segmentation and direct classification. Clinical studies demonstrate the high efficacy of such models in predicting outcomes in oncological, gastroenterological, hepatological, and other diseases, often outperforming traditional methods. However, implementation into routine practice is hindered by the lack of standardized cut-off values, variability in CT protocols, the problem of model interpretability (the "black box" issue), and a shortage of large annotated datasets. Future prospects are associated with the development of standards, the integration of automated tools into radiologists' workstations, and the conduction of prospective studies to prove clinical utility.
No takes yet. Share an insight, caveat, or question.
Tabakchi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: