Information Technology Reference
In-Depth Information
Fig. 5 Manual anatomical
landmark identi
cation on the
X-ray for the personalized 3D
reconstruction of the scoliotic
spine. Visible markers are the
identified landmarks on each
vertebra which are used in the
self-calibration of an X-ray
scene
sparse data with low redundancy [
7
,
47
], this generates multiple local minimums
when optimizing the non-linear equation system describing the radiographic
setup. Local correspondences also rely on the assumption that a point on an object
s
surface appears the same in the biplanar images in which it is visible. However due
to the phenomena exhibited by the X-ray modality, punctual point matches are not
necessarily a reliable matching feature for 3D bone reconstruction. For these rea-
sons, region-based comparisons via surface integration [
65
] may not only improve
the quality of 3D calibration results by incorporating additional data such as high-
level corresponding geometrical primitives (curves, surfaces), but can reduce the
number of degrees of freedom to solve the equation system.
Furthermore, the self-calibration of a biplanar X-ray system is a complex
mathematical problem dif
'
cult to solve. In order to improve the quality of current
self-calibration techniques [
7
,
28
] which rely on iterative algorithms optimizing the
retro-projection of sparse data points to solve a complex system of non-linear
equations, it is necessary to incorporate additional data into the system and to
