Computerized Medical Imaging and Graphics, Vol. 17, Nos. 415. pp. 25 l-256, 1993 Printed in the U.S.A. All rights reserved. 0895-61 I l/93 $6.00 + MI copyright 0 1993 Pergallon Prm Ltd. 3D DISPLAY OF HIGH RESOLUTION VERTEBRAL STRUCTURE IMAGES F. Peyrin *+ J.-P. Houssard*, E. Maurincomme*, G. Peix*, , R. Goutte*, A.-M. Laval-Jeantet*, and M. Amiel* *URA CNRS 1216, LTSU, INSA, 69621 Villeurbanne Cedex - France *Laboratoire de Contrhle Non Destructif par Rayonnements Ionisants - INSA- 1 Villeurbanne Cedex - France (Received 25 March 1993) Abstract-A methodology for three-dimensional (3D) representation of vertebral trabec&r structures was proposed. A set of X-ray CT images was obtained using a specific high resolution acquisition system. The images were then segmented in order to separate trabecuhu and cortical bone structures. Finally, the complex 3D surfaces were visualized using a volume rendering technique. Key W ords: X-ray tomography, 3D Display, Segmentation, Vertebral structures, tmbecular bone INTRODUCIION Precise measurements of bone structure parameters appear to be important in the evaluation of the bone strength, which considerably decreases in metabolic bone disease and osteoporosis ( 1). The current methods of bone densitometry, such as dual-energy X-ray ab- sorptiometry (DXA) and quantitative X-ray computed tomography (QCT), give only an approximation of bone mass or of bone density and, therefore, does not correlate perfectly with the fracture risk (2, 3). Recent studies suggest that the vertebral strength is not only dependent upon the bone density but also upon the architecture of the trabeculae (l-5). The structure of cancellous bone in normal and osteoporotic conditions has been the subject of current studies (1,4-6). The authors observed the anisotropy of the cancellous bone and proposed stereological methods for its measurement (7). The histological studies utilized microscopic images of parallel two-di- mensional (2D) slices ( 1,3,4) cut consecutively. In the best case, the sampling was both sag&al and coronal (6), or sag&al and transversal (5, 7). Parameters such as trabecular thickness, marrow space volume, and “star volume” (6) were defined. Multiple sections in- creased the statistical significance of the results but were not used for three-dimensional (3D) reconstruction. Similarly, clinical noninvasive studies proposed pro- tocols using computed tomographs involving volu- + Correspondence should be addressed to Dr. F. Peyrin, URA CNRS 1216. INSA, Batiment 502, 69621 Villeurbanne Cedex - France. 251 metric approaches (8, 9). A series of contiguous 5- 10 mm-thick sections provided 3D data (8). From these data, global parameters such as those given by histo- radiographic analyses, were computed. Nevertheless, vertebral trabeculae appear as very thin structures, most of them being 100 pm thick (7). Their measurement requires a high resolution method. Such an analysis is limited using conventional X-ray CT scanners. With the best commercially available CT scanners, the resolution is commonly limited to 250- 500 pm in the transversal plane and 1500 pm in the vertical direction. This is not sufficient to accurately observe the thin trabecular structures. In this paper, a methodology is proposed for 3D reconstruction of trabecular structure based on an ex- perimental high resolution X-ray CT scanner. Using this radiological CT device, which affords a spatial res- olution of around loo-150 pm, the aim was to rep resent and quantify directly the micro-architecture or- dinarily evaluated by histomorphometry. For this an acquisition device is described in the next section, and a description of the processing of a complete data set in order to obtain a 3D representation of the vertebral structure follows. DATA ACQUISITION A specific data acquisition system was used for the acquisition of slices of vertebra with a spatial resolution adapted to the structures to be studied (i.e., around 100-I 50 pm). The device consisted of an X-ray source, a linear array of CCD detectors, and a step motor al- lowing rotations of 0.05”.