Neural networks based approach to estimate body fat (%BF)

Abstract

Abstract: The amount of fat in human body composition relative to total body weight (%BF) is considered a determinant factor to a healthier and longer life. In this paper a neural network approach, that overcomes some of the current limitations of assessing %BF through skinfold thickness measurement with calliper devices, is presented. Neural networks recognised capabilities in modelling nonlinear problems can provide a valuable tool to deal with the inherent nonlinear behaviour of body tissues. The approach was tested on a sample of elder individuals, men and women, showing better performance when compared with two available alternative methodologies.