We present a new concept with mathematical background for the construction of a three-dimensional (3D) volumetric model of the human tibia based on two conventional orthogonal two-dimensional (2D) radiographic images. This approach is supported by a computer database containing a collection of 80 2D/3D image data sets of individual cadaveric tibiae. For each of these tibiae, the database contains digitized 2D orthogonal radiographic images in both anterior and lateral views, and the corresponding 3D CT data obtained by computerized tomography. To obtain a 3D model of a tibia for a given patient, shape matching is performed. The computer finds the most similar tibia to the patient's tibia among the 2D radiographic images in the database by applying a matching process. To improve accuracy, a 2D image warping procedure can be applied on the slices of the selected bone prior to 3D reconstruction. The warping process is controlled by the contour data of the two orthogonal views. We found that the 3D model thus achieved was useful for virtual preoperative planning and for simulation of the internal fixation of long bones.
Copyright 2001 Wiley-Liss, Inc.