BRAIN SURFACE RECONSTRUCTION FROM SLICE CONTOURS

Although surface-based analysis has become a standard tool in neuroimaging, existing surface reconstruction software is not fully general. For example, reconstruction of only a portion of the brain surface is not possible with existing tools, which make many restrictive problem-domain assumptions. Applications in which hand-tracing is the only reliable way of identifying tissue boundaries require new methods for surface reconstruction. Here, we present a set of software tools for creating surface representations from arbitrary image data. These tools provide a user interface designed to ease the process of hand-tracing brain structures of interest. A novel algorithm for surface reconstruction was developed to convert the hand-traced slice contour representation into a triangular mesh surface representation. Since hand identification of brain structures is laborious, we provided a user interface that addresses a range of editing requirements typical of this process. The intersection of the object surface with each slice image is traced, yielding a set of slice contours comprised of vertices with real-valued coordinates. The contour vertices need not be uniformly spaced, and can therefore sample the tissue of interest adaptively with its curvature. Since sub-surfaces are valid input, contours can be open or closed. Contour tracing is preferred over voxel labeling since tracing requires fewer locations on the object surface to be identified by the user. Tracing speed is increased by automatically timed contour vertex placement, allowing the user to quickly trace a feature while matching the density of contour vertices to curvature