Quantifying obesity from anthropometric measures and body volume data

Obesity has become a serious problem in several developed and developing countries. Three-dimensional photonic scanning (3DPS) is a useful tool to obtain accurate anthropometric measures and body volume data for body shape quantification. Some traditional models have been developed to estimate body fat percentages from anthropometric measures or body volume data for body composition classification and obesity quantification. However, these traditional models are very sensitive to the errors in anthropometric measures and body volume data. Small errors in anthropometric measures or body volume data reduces accuracy of body fat percentages estimated from 3DPS and may lead to misclassifications when quantifying levels of obesity. In this study, pattern recognition techniques, neural networks, were applied to develop a new model which can classify obesity levels from a combination of anthropometric measures and body volume data without estimating body fat percentages. The developed model and the traditional models were applied to determine 2209 male participants’ body composition classes for obesity quantification. The accuracy of the new and the traditional models was determined by comparing the estimated body composition classes with the real body composition classes obtained from dual energy X-ray absorptiometry scanning output. The results showed that the accuracy of the developed model was better than the traditional models. Therefore, the developed model provides more accurate results in body composition classification for obesity quantification.