Planar Procrustes analysis of tooth shape

Accurate quantification of variation in tooth shape is important in studies of dental development, which typically have involved measuring distances between subjectively identified landmarks, key points of correspondence on teeth. An established statistical framework now exists for the analysis of shape when objects are represented as configurations of landmark coordinates; allowing work with the full geometry of objects, which is otherwise lost. This approach was introduced here to the study of tooth morphology, demonstrating how after optimally matching shapes to account for the unwanted effects of location, scale and rotation, most standard descriptive and inferential statistical techniques can be adapted and applied successfully. The techniques are illustrated using a sample of buccal-surface images of central incisors from patients with hypodontia; a significant difference is found in mean buccal-surface shape (Hotelling's two-sample T(2)-test; P=0.004) when compared to a corresponding control group. Successful implementation of these methods depends on the accuracy and reliability with which the landmarks are collected; issues and problems to be addressed are discussed.