Elastic model-based segmentation of 3-D neuroradiological data sets

This paper presents a new technique for the automatic model-based segmentat ion of three-dimensional (3-D) objects from volumetric image data. The deve lopment closely follows the seminal work of Taylor and Cootes on active sha pe models, but is based on a hierarchical parametric object description rat her than a point distribution model, The segmentation system includes both the building of statistical models and the automatic segmentation of new im age data sets via a restricted elastic deformation of shape models, Geometr ic models are derived from a sample set of image data which have been segme nted by experts, The surfaces of these binary objects are converted into pa rametric surface representations, which are normalized to get an invariant object-centered coordinate system, Surface representations are expanded int o series of spherical harmonics which provide parametric descriptions of ob ject shapes. It is shown that invariant object surface parametrization prov ides a good approximation to automatically determine object homology in ter ms of sets of corresponding sets of surface points. Gray-level information near object boundaries is represented by 1-D intensity profiles normal to t he surface. Considering automatic segmentation of brain structures as our d riving application, our choice of coordinates for object alignment was the well-accepted stereotactic coordinate system. Major variation of object sha pes around the mean shape, also referred to as shape eigenmodes, are calcul ated in shape parameter space rather than the feature space of point coordi nates, Segmentation makes use of the object shape statistics by restricting possible elastic deformations into the range of the training shapes, The m ean shapes are initialized in a new data set by specifying the landmarks of the stereotactic coordinate system, The model elastically deforms, driven by the displacement forces across the object's surface, which are generated by matching local intensity profiles. Elastical deformations are limited b y setting bounds for the maximum variations in eigenmode space. The techniq ue has been applied to automatically segment left and right hippocampus, th alamus, putamen, and globus pallidus from volumetric magnetic resonance sca ns taken from schizophrenia studies. The results have been validated by com parison of automatic segmentation with the results obtained by interactive expert segmentation.