The aims of this work were to develop techniques for describing morphological variations
of shoulder bones and to test these on real datasets.
The robust measurement and description of anatomical geometry can provide accu-
rate estimation and better understanding of bone morphology. Feature lines were detected
automatically using crest line techniques and shape information from shoulder bones was
extracted based on the extracted feature lines. Redefinition of local coordinate systems
was proposed utilising the crest line technique.
Three dimensional statistical shape models (SSM) were built for a set of primate
humeri and scapulae. Two types of models were constructed: one incorporated the main-
tained original scale whilst the other used scaled bones. Variations were captured and
quantified by Principal Component Analysis (PCA). The application can be extended
generally to long bones and other complex bones and was also tested on human femora.
Techniques to predict the shape of one bone from its neighbour at a joint were
presented. PCA was used to reduce data dimensionality to a few principal components.
Canonical Correlation Analysis (CCA) and Partial Least Square (PLS) Regression were
applied to explore the linear morphological correlations between the two shoulder bones
and to predict the shape of one segment given the shape of the adjoining segment.