Spine Line Identification from Human Back Discretized Models by Geometric Differential Properties Analysis

The paper proposes a new recognition rule to automatically identify the spine line by detecting vertebral apophyses from discrete models of human backs collected by 3D scanning. Unlike previous 3D optical approaches, the process described here aims to directly detect spinal apophyses, following the posturologist’s method. These anatomical landmarks are identified not as areas of the back surface with specific shapes, but by searching for appropriate local shape perturbations. Except for vertebral prominences that are convex regions, spinal apophyses are generally unrelated to specific surface shapes. The rule has been tested on several human backs. For each, an experienced operator identified the spinal apophyses from palpation and indicated their position by applying an adhesive marker on the back surface. These positions are used as a reference to compare the vertebral apophyses automatically localized by the methodology proposed here. The experimental results show that the rule discriminates well even blurred vertebral prominences.