Computer-Assisted 3D Reconstruction of Serial Sections of Cortical Bone to Determine the 3D Structure of Osteons S. D. Stout, 1 B. S. Brunsden, 2 C. F. Hildebolt, 2 P. K. Commean, 2 K. E. Smith, 2 N. C. Tappen 3 1 Department of Anthropology, University of Missouri-Columbia, 107 Swallow Hall, Columbia, Missouri 65211-1440, USA 2 Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 South Kingshighway Blvd., St. Louis, Missouri 63110, USA 3 Department of Anthropology, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201, USA Received: 1 October 1998 / Accepted: 10 January 1999 Abstract. The objective of this study was to create three- dimensional (3D) images for the histomorphological study of osteons. Medical imaging technology was used to register digitized 2D images of serial decalcified histological sec- tions of bone, to segment the tissues of interest from the surrounding tissues, and to create 3D reconstructions from the segmented structures. Examination of the 3D recon- structions did not support suggestions in the literature that osteons have a spiraling organization. In contrast, the 3D reconstructions indicated that osteons have a complex pat- tern of organization that is dominated by branching. Exami- nation of the reconstructions also suggested that osteons described in the literature as being dumbbell shaped are actually artifacts of the plane of sectioning. This study dem- onstrated the applicability of imaging and visualization technology developed for the 3D reconstruction of medical images to the reconstruction of digitized 2D images of serial sections of bone and additionally demonstrated the feasibil- ity of using 3D reconstructions for the histomorphological study of osteons. The major microstructural feature of the compact bone of humans and most other large mammals and other verte- brates is the osteon (haversian system). It is generally un- derstood that osteons are produced by basic multicellular units (BMUs) of intracortical bone remodeling [1] and, al- though their two-dimensional structure has been well de- scribed, their three-dimensional (3D) structure remains un- certain. As with any biological structure, osteons exist in three dimensions, and to truly understand their nature, it is essential that we understand their three-dimensional mor- phology; but the mineralized composition of bone has made this task difficult. Tappen [2] describes a 3D study of resorption spaces and developing osteons using stained, decalcified, transverse se- rial sections of bone from several species (dogs, baboons, and humans). His findings, based upon tediously following histomorphological structures through contiguous sections, question some of the assumptions concerning the 3D struc- ture of osteons based upon the earlier work of Cohen and Harris [3] who also used serial decalcified sections of bone. A major problem in interpreting serial sections is one similar to that encountered with interpreting computed to- mography (CT) scan slices, i.e., being able to reconstruct individual two-dimensional cross-sections into a three- dimensional whole. Imaging and visualization technology developed for the 3D reconstruction of medical images [4–10] can be adapted to the reconstruction of digitized 2D images of serial histological sections of bone. This image processing methodology allows one to observe osteons three dimensionally, and store their images so that they can be analyzed, manipulated, and quantified. In this paper we report the results of a study for which medical imaging technology was applied to a series of decalcified serial sec- tions of dog bone to demonstrate the feasibility of perform- ing image processing and 3D reconstruction of osteons. Methods and Materials A series of black and white 35-mm photomicrographs of 34 se- quential contiguous transverse 30-m thick sections from a dog’s femur 1 from Tappen’s original research [2] were digitized at 16- bits and 1024 × 700 pixel resolution (Fig. 1). The original sections were stained with silver nitrate, and we optimized contrast by scanning from black and white 35-mm camera negatives, using a slide digitizer (Polaroid, Sprint Scan 35, CS-2700, Cambridge, MA). A SUN SPARC workstation (Sun Microsystems, Inc., Moun- tain View, CA) and ANALYZE™ software (Biomedical Imaging Resource of the Mayo Clinic, Rochester, MN) were used for all image processing and image generation. ANALYZE™ software is specifically designed for multidimensional imaging and processing and contains the most complete set of processing tools currently available for these tasks [6, 8]. Histogram matching was employed to correct for variations in exposure, processing, and digitization among images (Fig. 2a). The resolution of the images was then reduced to 6-m pixels from 3-m pixels and the gray scale changed from 16 bits to 8 bits. All 34 slices (sections) were registered (Fig. 2c). The area of interest was reduced to a common field of view. An adaptive histogram process was used to reduce the effect of nonuniform illumination and to enhance the bone structures (Fig. 2e). Addi- tional processing was performed with spatial filters, edge detec- tion, and morphological operators to segment and create schemat- ics of specific structures (Fig. 2f). We were able to view a 3D block of bone from various angles and as a transparent object. Results 3D volume rendering and sectioning at a number of differ- Correspondence to: S. Stout 1 The sections were part of a series of 180, 30-m thick, decalci- fied serial sections from an earlier study by Dr. N. Tappen (1977). Calcif Tissue Int (1999) 65:280–284 © 1999 Springer-Verlag New York Inc.