Information Technology Reference
In-Depth Information
wedging), we present here the components of a hybrid statistical and image-based
biplanar reconstruction method [
26
]. The spine centerlines extracted from the
pre-operative images are used to map the 3D reconstruction of the spinal curve in a
low-dimensional representation of a scoliotic database, and perform a statistical
modeling of the anatomy based on an analytical regression. The model is re
ned
locally at each vertebral level via a segmentation method based on a level set
surface evolution paradigm.
4.4.1 Training Data
The statistical model was built from a 3D database containing 711 scoliotic spines
demonstrating several types of deformities. Each scoliotic spine in the database was
obtained from biplanar stereo-reconstructions. It is modeled with 12 thoracic and
5 lumbar vertebrae (17 in total), represented by 6 landmarks on each vertebra
(4 pedicle extremities and 2 endplate center points).
Segmentation of the scoliotic vertebrae on the X-ray images was performed by
using generic vertebra priors obtained from serial CT-scan reconstruction of a
cadaver specimen (Fig.
9
). Models were segmented using a connecting cube algo-
rithm [
42
] with 1-mm-thick CT-scan slices taken at 1-mm steps throughout the dry
spine. The atlas is composed of 17 cadaver vertebrae (12 thoracic and 5 lumbar). The
atlas is divided into 3 levels of polygonal mesh catalogues of increasing complexity,
to adopt the widely used multi-resolution registration approach where coarse-to-
ne
geometrical models are applied for optimal convergence, where models were
composed between 3,831 and 6,942 vertices depending on the vertebra level. The
same six precise anatomical landmarks (4 pedicle tips and 2 on the vertebral body)
were annotated on each individual model.
Fig. 9 a Atlas of 17 vertebral models obtained from computer tomography (CT) and represented
with 3D Fourier descriptors. b Annotated landmarks of the
first lumbar vertebra
