Evaluation and Enhancement of a Procedure for Generating a 3D Bone Model Using Radiographs

Volumetric information about the patient's anatomy is quite valuable for medical diagnosis. Computed tomography (CT) is the common imaging modality for 3D visualization of bone tissue but rising costs in health care system demand for new approaches. A promising one is to use a 3D model being deformable under the constraint of statistical plausibility. The model is adapted to the patient's anatomy by extracting the specific bone features from several conventional radiographs (2D--3D registration). These have to be acquired under different angles thus providing the features' 3D position by means of which the model is deformed. The resulting bone representation may then be used for medical diagnosis instead of using CT data. Present work validates accuracy of the resulting bone shape and thus of the diagnosis relying thereon. Results are starting point for further implementations and modifications in order to reduce remaining errors.