Toward quantitative characterization of the caudate nucleus through CT image enhancement.

ABSTRACT

Dilated lateral cerebral ventricles have been reported in association with psychiatric disorders. The dilatation is likely secondary to changes in anatomically related centers in the brain. Our attention is focused on the caudate nucleus and the need to develop quantitative characterization of it in psychiatric research. A computer method is described for accurately detecting the edges of the caudate nuclei in CT scans. The method consists of a two-dimensional filtering step to reduce image noise, followed by gradient computation for edge enhancement, and then an edge detection step to identify and record the caudate nuclei cross sections. Results from both head scans and synthetic images indicate that the method is successful even for low-contrast edges where the standard deviation of image noise approaches the difference in mean HU across the edge. The availability of an automated method of delineating the caudate nucleus permits the cross-sectional area and shape to be determined, the density to be estimated, and, from a set of closely spaced slices, volume information to be obtained.