Using parts and geometry models to initialise Active Appearance Models for automated segmentation of 3D medical images

In recent years, statistical shape models, of which Active Appearance Models (AAMs) are a subset have been increasingly applied to the automatic segmentation of medical images. AAMs are a local search technique requiring good initialisation. In 3D automatic initialisation can be achieved by multiple initialisations, registration, template matching or by application dependent heuristics. The first three can be sub-optimal in certain situations, whilst the last is not generic. We describe a generic, fast and automated method of initialising 3D AAMs using sparse local models of texture (the parts) together with a graph capturing their pairwise geometric relationships. Initialisation then becomes a matter of searching for the parts using the parts-and-geometry model, from which the necessary pose and shape parameters are obtained. We demonstrate the method by applying it to the segmentation of 10 subcortical structures from 3D MRI sequences of the head.