A comprehensive regression tree to estimate body weight from the skeleton.

The purpose of this research is to estimate actual body weight (in particular obesity) from the human skeleton. Known individuals (total n=187) were studied from the Bass Collection from the University of Tennessee, Knoxville. This research combines bone density, cross-sectional geometry of the femur and skeletal pathologies. Bone mineral density was calculated for the proximal femur. Three-dimensional bone surface models were created from computed tomographic scans. Cross-sectional geometry of the femur was calculated at five locations along the diaphysis. The pathologies analyzed were heel spurs, diffuse idiopathic skeletal hyperostosis (DISH), and tibial osteoarthritis. The best regression tree model included only four variables. The first split to estimate body weight was the minimum moment of inertia (I(y) ) at the distal femur, the second was midshaft width, then anteroposterior thickness at the proximal slice and the final split was the degree of DISH (SD 17.1-31.0 kg). The ability to estimate body weight from the skeleton is one more useful tool for the osteologist.