Multimodal imaging techniques in sport

Anthropometry and body composition analysis have been largely applied in scientific research to identify different subgroups within a certain population. Recent progress in both fields is changing the way these study are performed and open a lot of new possibilities. Body composition analysis can exploit imaging devices like dual-energy X-ray absorptiometry (DXA) that currently represents the gold standard to estimate fat mass, lean mass, and bone mineral content and new generations of acquisition devices like ultrasound scanners, air displacement plethysmographs, etc. Anthropometry is then becoming fully digital: 3D scanners can now acquire very accurate models of human subjects, allowing automatic or computer assisted measurement of body lengths, breadths, diameters, volume and surface area. Considering that current techniques used to estimate body surface area as well as total body or segmental volume are based on the use of formulae, it is easy to understand the potential improvements provided by digital models. A relevant task for research in body shape and composition analysis is therefore related to the validation of digital measurements. In our work we developed computer tools to evaluate a variety of shape parameters from original noisy scanner models, trying to validate them following two main research directions. The first consists of comparing body scanner data with other well-established biometric technologies. We therefore compared measurements obtained with standard tape-based procedures with similar ones simulated with a custom made software tool on the digital models, obtaining comparable results for most of the analyzed measures. Results indicate that digital anthropometry allows for accurate and reproducible measurements also when operators lack specific training. The second research approach aims at developing a totally automatic analysis of body shape in order to obtain measurements from raw scanner data. Given the great amount of data made available by the body scanning technology, developing viable, automatic and robust procedures is essential. Computational geometry tools like remeshing and skeletonization algorithms, etc. have been applied to analyze the models and automatically extract from them shape parameters independently on pose variations and with sufficient robustness against acquisition noise or holes. Being these parameters not directly comparable with tape-based methods, a possible validation approach we are testing is related to a direct correlation of the measurements with composition data acquired with the DXA scanner. In a preliminary study we found that several automatically extracted body measurements (obtained in the abdominal region) are highly correlated with trunk fat (r≥0.9 with maximum Anterior-Posterior diameter, maximum Mean Diameter and maximum of slice Area) and total body fat (r=0.76 with maximum Anterior-Posterior diameter, r=0.89 with maximum Mean Diameter, r=0.88 with maximum of slice Area).