Computerized three-dimensional volume analysis from magnetic resonance images for characterization of brain disorders

Quantitation of regional brain function as a means for the characterization of brain disorders is discussed. Such quantitation can be accomplished by three-dimensional volume analysis based on magnetic resonance images, even though simple thresholding cannot be used for the segmentation of such images. As a first step toward such analysis, techniques for identifying and isolating the contents of the skull and, in particular, cerebrospinal fluid are described. The techniques of gradient-based boundary tracking, partitioning in a two-dimensional feature space, and shape-based interpolation appear to be effective for achieving this limited aim.<
>