The following steps are used to create a population average from a set of input specimen based on 3D volumetric data. In the example below,
3D anatomic landmark points placed on three input specimen guide the process to create a population average.
|Input specimen. Volume data (CT, MRI, etc.)
with surfaces and a homologous set of
3D landmark points are placed on
each specimen. Landmark points should
be concentrated around features of
|The 3D landmark point sets are used to
represent the population. Orientation and
scale may be different among specimen.
|Specimen landmark point sets undergo rigid alignment
to remove orientation differences. Scale differences
can be maintained or removed as desired.
|Average pointset is computed using
Generalized Procrustes Alignment (GPA) to
remove scale (if desired). Resulting in the
target average pointset (orange dots).
|Each scan volume is warped to the average pointset
using Thin-Plate Spline (TPS) warping. All deformed
volumes are then merged together.
|The resulting population average volume can
be manipulated just as any of the original specimen
volumes. Surfaces can be extracted, analyzed,
and rendered from this average volume.