A Method for Detection and Modeling of the Human Spine Based on Principal Curvatures

The detection and modeling of the human spine from scanned 3D data is an important issue in biomedical shape analysis. It can be useful for avoiding invasive treatments like radiographs, taken for the purpose of monitoring spine deformations and its correction, as is the cases in scoliosis. This is especially important with children. This work presents a new method for the detection of the human spine from 3D models of human backs formed by triangular meshes, and taken with a range sensor. The method is based on the estimation of the principal curvatures directions, and by joining valley points along these directions. Results are presented with the method applied to scanned 3D models of real patients.