The authors present a computerized system for reconstructing the 3-D structures of pulmonary blood vessels from a set of X-ray CT images. The system uses a knowledge base in which anatomical knowledge and general rules concerning the way blood vessels extend are acquired. Others have already done some work on the automatic 3-D reconstruction of organs, but the authors have been analyzing ordinary chest X-ray CT images with 1-cm slices, and the object organs, pulmonary blood vessels, are slender. The system combines AI approaches with image-processing technologies.<<ETX>>