Artificial Neural networks Models Fail to Accurately Estimate Fat and Lean Body Mass

Research Question: Can artificial neural network-based Body Composition Prediction (BOMP) models provide valid estimates of fat mass (FM) and lean body mass (LBM) compared with dualenergy X-ray absorptiometry (DXA)?Key Findings: In this cross-sectional study of 641 young adults (303 females, 338 males), BOMP models showed high correlations with DXA (R = .81-.95), except for female LBM (R = .66).However, wide limits of agreement (>20%), failure to meet error thresholds, and inconsistent equivalence with DXA indicated poor validity.Overall, BOMP models were not accurate for FM and LBM estimation, particularly at the individual level.