AUTHOR=Graybeal Austin J. , Compton Abby T. , Swafford Sydney H. , Brandner Caleb F. , Johnson Molly F. , Kaylor Maria G. , Haynes Hunter , Stavres Jon 

TITLE=A predictive model for body water and fluid balance using 3D smartphone anthropometry

JOURNAL=Frontiers in Physiology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2025.1577049

DOI=10.3389/fphys.2025.1577049

ISSN=1664-042X

ABSTRACT=BackgroundBody fluid volumes, including total body water (TBW), extracellular fluid (ECF), and intracellular fluid (ICF), are crucial indicators of body composition, and the distribution of these fluids is essential for assessing hydration status and fluid accumulation. Although fluid volumes are commonly measured with bioelectrical impedance devices, several challenges hinder the application of this technique. However, 3D smartphone scanning applications that automate body volumes and other anthropometric estimates may provide a viable alternative to body fluid assessments.MethodsA total of 338 participants underwent fluid volume assessments using bioelectrical impedance spectroscopy (BIS) and collected body volumes and anthropometric data using a 3D smartphone scanning application. Then, LASSO regression was used to develop new TBW and ECF prediction model in a subset of participants (n = 272), which was subsequently tested in the remaining participants (n = 66). Smartphone-derived ICF was calculated as the difference between smartphone-predicted TBW and ECF. Fluid overload and imbalance were determined using ECF/TBW and ECF/ICF, respectively, and subsequently predicted from the retained variables using receiver operating characteristic curve analyses and logistic regression.ResultsEstimates from each of the newly-developed prediction models were not significantly different from the estimates produced using BIS (all p ≥ 0.70) and revealed acceptable agreement (TBW: R2 = 0.91, RMSE = 3.24 L; ECF: R2 = 0.94, RMSE = 1.10 L; ICF: R2 = 0.87, RMSE = 2.29 L) when evaluated in the testing sample (n = 66), although proportional bias was observed (p < 0.001). Smartphone-predicted fluid overload (AUC: 0.81 [95%CI: 0.70, 0.92]; sensitivity + specificity: 1.53 [95%CI: 1.39, 1.67]) and imbalance (AUC: 0.76 [95%CI: 0.64, 0.88]; sensitivity + specificity: 1.40 [95%CI: 1.24, 1.56]) demonstrated acceptable diagnostic performance.ConclusionSmartphone scanning applications can accurately assess body fluid volumes and imbalances, presenting new possibilities for health screening beyond clinical environments.