Noninvasive assessment of bone architecture by magnetic resonance micro-imaging-based virtual bone biopsy

The mechanical competence of trabecular bone, the type of bone constituting the vertebrae and ends of the long bones, is largely determined by the boneʹs volume fraction and architectural makeup. Experimental and clinical evidence suggests that up to 50% of trabecular bone strength is determined by parameters characterizing the networkʹs architecture. Although rarely clinically indicated because of its invasiveness, a bone biopsy can provide detailed insight into the boneʹs structural arrangement. Recent advances in magnetic resonance micro-imaging ((mu)MRI), in conjunction with new image processing and feature extraction approaches, now allow detailed structural information to be obtained by what we refer to as "virtual bone biopsy" (VBB). Governed by signal-to-noise, the voxel size achievable in vivo is of the same order as trabecular thickness (100-150 (mu)m). Therefore, methods had to be conceived to deal with the inherently fuzzy nature of the objects to be extracted. It is shown that a cascade of image processing steps, starting with noise deconvolution of the images to yield bone volume fraction maps, followed by what we refer to as "subvoxel processing" for resolution enhancement, are essential for evaluating topology and scale of the trabecular network. Digital topological analysis of the binarized and skeletonized images can provide a detailed picture of network connectivity and the nature of the structural elements (plate and strut architecture). In addition to its complicated topological makeup, trabecular bone is highly anisotropic since bone adapts to the stresses to which it is subjected (Wolffʹs law). Therefore, evaluation of structural orientation or fabric is of interest as well. Several methods are discussed to quantify structural anisotropy, including digital topological analysis. Although connectivity is an important determinant of the boneʹs mechanical behavior, the thickness of the structural elements is equally relevant. Here again, the limited spatial resolution achievable in vivo precludes the use of the classical histomorphometric approaches for thickness measurement. A method conceived recently in the authorsʹ laboratory is the fuzzy distance transform, which obviates the need for binarization and which is shown to - provide accurate measurements of structural thickness. Excerpts are provided from applications of the VBB to the study of postmenopausal osteoporosis, male hypogonadism, and secondary hyperparathyroidism. Finally, the serial reproducibility achievable suggests the VBB to be suited for assessing treatment efficacy longitudinally.