Feature Recognition Using Volume Data with Application Spinal Fracture Diagnosis

Abstract

This paper describes feature recognition methods for the evaluation of 3D geometry of vertebral bones and spinal anatomic curve in the diagnosis of compression and burst fractures. The method uses a radial B-spline curve to extract the feature of the ellipse-like vertebral body on a transverse section in which a concave feature is also recognized to evaluate the compression of the canal. 3D feature of infers the anatomic curve of a vertebral body is recognized by linearly regressing the centers of B-spline approximate ellipse-like boundaries of the transversal sections passing the vertebral body. Then, the reduced angle and height for recovering the compression fracture can be calculated by comparing the regressed centerlines of neighboring bodies of the fracture body with the normal spinal anatomic curve. The prototype system can be used as a qualitative and quantitative tool for the diagnosis of compression and burst fractures using transverse sections, and for the instruction to plan accurate surgical procedures.