FW: GEN RE: Types of data

[Don Kirkup]

> but ovate is rather difficult to parameterize. It may be possible to
> get ellipsoidal parameters, but width and length are clearly
> insufficient.
>
> Does anybody know about good parameterization of shapes, so they can
> be stored in more objective forms?

Wouldn't the most objective representation of shape be given by a set of
coordinates for points placed around the outline? If orientation of the
object is important, then give landmark coordinates too eg. mark the organ
base and apex. From that info, parameters such as object area, perimeter
length, circularity (perimeter length*perimeter length/area), centroid,
longest and shortest axes, aspect ratios etc. can be calculated.

You'd also be able to standardise your outlines (eg. using Bookstein's
transformation) and compare against reference shapes (eg elliptic fourier
analysis), pairwise comaprisons using (thin plate splines) or fit to average
shape using Procrustes metric.

Feature extraction from digital images (eg defining the object outline,
sampling of points around outline)and calculation of the shape parameters
and export to spreadsheet is virtually automatic in packages like OPTIMAS.
James Rolf's NTsysPC will perform the basic standardisations, thin plate
splines and EFA.

don