Measurement of carpal bone geometry in CT images

Contiguous CT (computed tomography) slice images of the carpal region of human cadaver extremities were acquired. The boundaries of the individual carpal bones were located in the images, using a boundary tracker that uses intensity, gradient, angular, and neighborhood information to evaluate alternative boundary locations. A dynamic programming technique is used to find an optimal path through a region of the image containing an estimated boundary location defined by a template. Boundaries are found for the eight carpal bones, the proximal end of the five metacarpals, and the distal end of the radius and ulna, in each section in which each bone appears. The contours of each bone are grouped together and, assuming that the contours constitute a set of points representing the surface of the bone, the dimensions, volume, surface area, and orientation of the bones are determined. The distances between the bones are computed. A tiled surface is computed connecting the adjacent planar contours, and a rendering is generated of the resulting 3-D structure. Data were collected on six wrists, and the measurements of size and volume for the carpal bones from these specimens were compared with direct measurements of the bones after dissection. The image-based measurements agreed closely with the direct measurements.<>