ABSTRACT
In this pilot study, we developed a set of computer vision based surface segmentation and statistical shape analysis algorithms to study genetic influences on brain structure in a database of brain MRI scans of normal twins. A set of manually delineated 3D parametric surfaces, representing the lateral ventricles, was deformed, using a Navier-Stokes fluid image registration algorithm, onto all the images in the database. The geometric transformations thus obtained were used to propagate the segmentation labels to all the other images. 3D radial distance maps were derived to encode anatomical shape differences. The proportion of shape variance attributable to genetic factors, known as the heritability, was estimated from the shape models using a restricted maximum likelihood method to increase statistical power. Segmentation errors associated with projecting labels onto new images were greatly reduced through multi-atlas averaging. The resulting algorithms provide a convenient and sensitive tool to recover and analyze small intra-pair image differences, and will make it easier to detect genetic influences on brain structure.
