The objective of this paper is to automatically detect the back valley on a polygonal mesh of the human trunk surface. A 3D camera system based on the projection of a structured light is used for the acquisition of the whole trunk of scoliotic patients. A quadratic fitting method is used to calculate the principal curvatures for each vertex. It was determined that 3 levels of neighbors were sufficient to detect the back valley. The proposed method was evaluated on a set of 61 surface trunks of scoliotic patients. The results were validated by two orthopedic surgeons and were estimated to 84% of success in the automatic detection of the back valley. The proposed method is reproducible and could be useful for clinical assessment of scoliosis severity and a non-invasive progression follow-up.
展开▼