High-resolution Computed Tomography for Clinical Imaging of Bone Microarchitecture

Abstract

Background

The role of bone structure, one component of bone quality, has emerged as a contributor to bone strength. The application of high-resolution imaging in evaluating bone structure has evolved from an in vitro technology for small specimens to an emerging clinical research tool for in vivo studies in humans. However, many technical and practical challenges remain to translate these techniques into established clinical outcomes.

Questions/purposes

We reviewed use of high-resolution CT for evaluating trabecular microarchitecture and cortical ultrastructure of bone specimens ex vivo, extension of these techniques to in vivo human imaging studies, and recent studies involving application of high-resolution CT to characterize bone structure in the context of skeletal disease.

Methods

We performed the literature review using PubMed and Google Scholar. Keywords included CT, MDCT, micro-CT, high-resolution peripheral CT, bone microarchitecture, and bone quality.

Results

Specimens can be imaged by micro-CT at a resolution starting at 1 μm, but in vivo human imaging is restricted to a voxel size of 82 μm (with actual spatial resolution of ~ 130 μm) due to technical limitations and radiation dose considerations. Presently, this mode is limited to peripheral skeletal regions, such as the wrist and tibia. In contrast, multidetector CT can assess the central skeleton but incurs a higher radiation burden on the subject and provides lower resolution (200–500 μm).

Conclusions

CT currently provides quantitative measures of bone structure and may be used for estimating bone strength mathematically. The techniques may provide clinically relevant information by enhancing our understanding of fracture risk and establishing the efficacy of antifracture for osteoporosis and other bone metabolic disorders.

Introduction

The skeleton is composed of cortical and trabecular bone, both contributing to bone strength and the resistance of bone to fracture. The strength of bone and risk of fracture are important outcomes in the study of growth and peak bone accrual, aging, postmenopausal bone loss, cancer-related bone loss, and for conditions such as diabetes, osteogenesis imperfecta, osteoarthritis, rheumatoid arthritis, and others. Typically, clinical assessment of skeletal health is based on measures of bone mineral density (BMD), usually obtained using dual-energy xray absorptiometry (DXA), a two-dimensional, projection-based radiographic technique that measures integral BMD of both cortical and trabecular bone (areal BMD). In addition to DXA, three-dimensional quantitative CT (QCT) is used to assess BMD. This three-dimensional technique measures volumetric BMD and permits characterization of bone geometry and density as elements of fracture risk. Furthermore, QCT can examine cortical and trabecular bone independently.

BMD only explains about 70% to 75% of the variance in strength [2], while the remaining variance is due to the cumulative and synergistic effect of factors such as bone macro- and microarchitecture, tissue composition, and microdamage [30, 132]. In a multicenter fracture intervention trial, the antifracture efficacy of all drugs tested was only partially explained by their effects on BMD [9]. In this context, and specifically in osteoporosis, the concept of bone quality emerged [132]. Bone quality represents different properties of bone to researchers and could encompass one or all of the factors mentioned. Trabecular and cortical bone play important roles in the prediction of bone strength and are affected by age, gender, and metabolic conditions and have varying responses to therapy. Sites containing predominantly trabecular bone, such as the hip, spine, and wrist, are most frequently associated with increased fracture risk [39]. Traditionally, trabecular microarchitecture is assessed from bone biopsies by two-dimensional histomorphometry [116], applying stereologic principals to measure trabecular bone volume fraction (the ratio of mineralized bone volume to total volume [BV/TV]), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N). On the other hand, cortical bone constitutes about 80% of total bone mass. In previous studies, cortical thinning [119] and increased cortical porosity [13] were important factors in the assessment of osteoporosis and bone strength. Recently, focus increased on the ultrastructure of cortical bone that can be attributed to resorption spaces, merging of Haversian canals, and clustering of osteons [6, 77]. Since the ultrastructure of cortical bone has a major impact on its mechanical properties [131, 152], characterizing cortical ultrastructure is also important in the context of bone strength and prediction of fracture risk. With the advent of improved three-dimensional imaging techniques, such as micro-CT [51, 126], high-resolution peripheral QCT (HR-pQCT) [14, 82], and multidetector CT (MDCT) [73], it is possible to perform in vitro and in vivo imaging of bone across different structural scales from the whole bone to the ultrastructural level.

In a recent publication for the radiographic imaging communities, we reviewed three-dimensional techniques for assessing bone structure in osteoporosis [86]. We described MRI, image processing, and CT, concluding these modalities have the potential to play an important role in imaging three-dimensional trabecular microarchitecture in osteoporosis. In this review, we focus on the development and application of high-resolution CT for quantifying cortical and trabecular bone structure covering specific clinical applications of interest to the orthopaedic research community.

We reviewed (1) the fundamentals of high-resolution CT for evaluating trabecular microarchitecture and cortical ultrastructure, (2) use of micro-CT for studying bone specimens ex vivo, (3) use of MDCT for in vivo human imaging studies, and (4) recent studies using HR-pQCT to characterize bone structure in the context of skeletal disease, particularly its ability to discriminate between subjects with and without fractures and monitor longitudinal response to therapeutic intervention.

Search Strategy and Criteria

To determine the relevant articles, we used the PubMed (PM) and Google Scholar (GS) search engines. The following list of search phrases was used, with the number of results reported parenthetically: “CT” AND “trabecular bone microarchitecture” (PM: 14; GS: 549); “microCT iliac crest biopsy” (PM: 17; GS: 122); “microct” AND “cortical bone porosity” (PM: 19; GS: 74); “XtremeCT” (PM: 8; GS: 139); “HR-pQCT” (PM: 68; GS: 247); “MDCT” AND “bone structure” (PM: 7; GS: 111); “SR-μCT” AND “bone structure” (PM: 8; GS: 82). From these, we selected those articles we believed most relevant.

Fundamentals of CT Imaging

CT is a three-dimensional radiographic imaging technique. The image formation process begins with the acquisition of sequential radiographic projections captured over a range of angular positions around the object of interest. The cross-sectional field of view is reconstructed using established computational techniques based on radon projection theory [50]. Similar to simple radiography, the reconstructed image’s intensity values represent the local radiographic attenuation: a material property related to the object’s electron density (atomic number and mass density). The contrast between soft and mineralized tissue in CT is high, due to the relative electron-dense inorganic component (calcium hydroxyapatite) of the bone matrix [8]. Since the logarithm of the measured absorption scales linearly with the length of material the beam has penetrated, simultaneous quantitative measurements of bone density are possible. Calibration of grayscale linear attenuation to BMD is accomplished by imaging reference phantoms containing objects with known hydroxyapatite concentrations [21, 49].

These principles capture high-resolution images of bone across a range of structural scales. Several classes of CT devices are presently used for high-resolution imaging of trabecular microarchitecture and cortical ultrastructure (Table 1). Techniques with spatial resolution between 1 and 100 μm are referred to as micro-CT and may replace tedious serial staining procedures required by histomorphometric analysis of thin sections and offer the possibility of longitudinal in vivo investigations in small animals, such as mice and rats. Many early micro-CT approaches used synchrotron radiation (SR) [59], which is still the method of choice for ultrahigh-resolution applications. The use of desktop laboratory scanners equipped with xray tubes is much more convenient than performing an experiment at one of the few synchrotron facilities available worldwide. Initial and ongoing university-based research in the past decade has led to the development of a variety of commercial xray tube-based micro-CT scanners (Table 1).

Table 1.

Descriptive summary of high-resolution CT technologies available

Modality	References	Primary manufacturers	Skeletal sites	Field of view size (mm)	Voxel size (µm)	Effective dose	Typical scan time
Micro-CT	55, 129, 135, 151	GE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland) Siemens (New York, NY) SkyScan (Kontich, Belgium) Xradia (Pleasanton, CA)	Specimens Biopsies (ex vivo)	2–100	0.3–100 (isotropic)	NA	30 minutes to 3 hours
MDCT/fp-vCT	5, 41, 43, 44, 69, 70, 73, 85, 118, 124	GE Heathcare (Waukesha, WI) Philips (Amsterdam, The Netherlands) Siemens (New York, NY) Toshiba Corp (Tokyo, Japan)	Specimens (ex vivo) Spine Femur Forearm (in vivo)	100–250	156–300 (in plane) 300–500 (slice thickness)	0.1–5 mSv	< 30 seconds
HR-pQCT	14, 82, 130	Scanco Medical AG (Brüttisellen, Switzerland)	Specimens (ex vivo) Distal radius Distal tibia(in vivo)	126	41–123(isotropic)	3–4 µSv	3 minutes

NA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.

Application of standard whole-body MDCT to imaging trabecular bone in the central and peripheral skeleton was investigated at several research institutes [5, 43, 58, 70, 73]. Compared to standard two-dimensional QCT, volumetric MDCT imaging studies use higher-dose acquisition protocols (3 mSv versus 0.06–0.3 mSv) with a higher in plane resolution (200–300 μm versus 500–1000 μm) and smaller slice thickness and spacing (500 μm versus 1–10 mm). Recently, a standard MDCT gantry was combined with two-dimensional flat panel detector technology to provide rapid continuous acquisitions at high isotropic spatial resolution [60, 124]. In the last 5 years, a high-resolution, limited-field-of-view CT device became commercially available for dedicated imaging of bone structure in the peripheral skeleton [14, 80, 82, 97]. The HR-pQCT imaging system consists of a microfocus xray source and high-resolution charge-coupled device (CCD) detector that can produce tomographic images with a nominal resolution as high as 41 μm for a 12.6-mm field of view.

Micro-CT

Micro-CT has achieved widespread use in the laboratory for rapid, nondestructive imaging of bone specimens [47, 51, 108] and noninvasive imaging in animal models [54, 148]. The pervasive use of this technology at many research institutes invested in bone science research has widely eclipsed traditional histomorphometry for evaluating bone microarchitecture. However, micro-CT does not provide direct information on cellular function and remodeling activity, which continues to be the domain of bone histology.

Conventional laboratory micro-CT typically uses a cone-beam, polychromatic xray source, which produces photons spanning a broad range of energies. In contrast, SR micro-CT, only available at a limited number of particle accelerator facilities worldwide, is typically performed using a parallel, monochromatic beam [72, 83, 127, 141]. While the first commercial micro-CT scanner consisted of a single-row 512-pixel detector [126], modern scanners employ areal CCD detectors up to 11 megapixels and are capable of acquiring projection data for more than 1000 slices simultaneously [129, 135, 151]. Dedicated specimen ex vivo scanners are typically designed with the specimen oriented on a high-resolution motorized stage for translation and rotation. In this scenario, the source and detector remain in a fixed position during the scan, while the specimen rotates in the field of view. Conversely, preclinical micro-CT systems designed for in vivo imaging of small animals utilize a fixed gantry with the xray source and detector rotating and translating about the field of view [54, 147].

Analogous to clinical QCT, the grayscale attenuation values of reconstructed micro-CT images can be converted to hydroxyapatite concentration. Various calibration procedures based on idealized phantoms or theoretical calculations were established to derive relations between attenuation and BMD [21, 72, 102, 107, 110, 113]. However, micro-CT imaging is subject to xray scattering (SR and polychromatic micro-CT) and beam-hardening effects (polychromatic micro-CT) that can introduce nonnegligible error in the depiction of mineralization that is spatially variant and dependent on the geometry and composition of the object imaged [48, 79, 102]. Methods to minimize beam-hardening effects in conventional micro-CT include xray filtration to block “soft” low-energy xrays [102] and empirical corrections based on phantom measurements [21]. Using these procedures, apparent BMD and tissue level mineral density can be accurately measured (r² = 0.78–0.99) in specimens of similar size and composition [21, 79].

Morphometric indices analogous to classical histomorphometry can be calculated from micro-CT images of trabecular and cortical bone. Comparison of structural parameters of specimens scanned with these systems and mechanical testing suggest the amount of bone and the architecture of trabecular bone contribute to mechanical strength [57]. Advanced image-processing methodologies are used to quantify trabecular bone microarchitecture beyond measures of bone volume fraction (BV/TV). Specifically, direct three-dimensional measures of mean distances and measures of structural heterogeneity are used to characterize trabeculae and marrow spaces [64, 65] (Fig. 1). The degree of anisotropy, a measure of the degree of structural orientation of the trabecular network, can be calculated from the principal structural directions calculated by the mean intercept length techniques [63] and is highly related to the directional dependence of bone’s biomechanical properties [115]. A measure of the structural connectedness was also adapted to bone, based on the Euler number [114]. The shape of the trabecular structure is characterized using the structure model index (SMI), an index of surface convexity that estimates the degree to which the structure consists of rod-like or plate-like elements [66]. Furthermore, several groups developed algorithms to decompose the trabecular structure to independently quantify the volume and scale of rod-like and plate-like elements [92, 117, 139].

Fig. 1A–B.

Three-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].

The ultrastructure of cortical bone is an important determinant of bone strength [128, 131], critical in fracture initiation and propagation [146], and known to change with age [37], disease [7], and therapy [11]. Volumetric and morphologic characterization of the cortical ultrastructure has predominantly focused on Haversian and Volkmann canal network of cadaveric femoral neck specimens [13, 32, 36, 38, 153]. Resolution improvements heralding the evolution of nano-CT (CT with submicron resolution) recently paved the way for a complete evaluation of cortical bone ultrastructure, including the distribution of osteocyte lacunae [131, 146].

While volumetric density and microarchitecture information provide improved fracture risk prediction and some explanation for treatment efficacy, more direct estimates of bone mechanical strength that inherently account for geometry, microarchitecture, and even composition are the ultimate goal for improving fracture risk prediction and management of osteoporosis. Computational modeling approaches were introduced to take advantage of the detailed information in high-resolution images of bone. Finite element analysis (FEA) is a common computational tool in engineering fields, critical to design and failure analysis. Applied to high-resolution images of bone, the apparent biomechanical properties (stiffness, elastic modulus) of a biologically complex microstructure are computed by decomposing the structure into small cubic elements (the voxels) with assumed mechanical properties [109, 143] (Fig. 2). Numerous studies have utilized micro-FEA techniques to investigate the micromechanics of bone strength, failure, and relation to bone microarchitecture [81, 142].

Fig. 2.

A micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.

As an ex vivo imaging modality, the application of micro-CT to clinical research is limited to examining small bone biopsy specimens. There are numerous data over the last 20 years characterizing age, gender, and anatomic differences in cadaveric specimens of bone [13, 45, 52, 64, 67, 68, 94]. Clinical applications generally preclude access to the most relevant sites, such as the spine and proximal femur. Accordingly, a minimally invasive bone biopsy is typically acquired from the iliac crest [75]. Trabecular microarchitecture at the iliac crest reflects vertebral fracture status [56, 74] and changes after the onset of menopause [1]. Quantification of bone structure from iliac crest biopsies is also an important end point in longitudinal drug efficacy studies of parathyroid hormone [33, 53, 76, 121], strontium ranelate [3], and various bisphosphonates [11, 12, 46, 112, 120, 122, 123]. Borah et al. [11] recently reported major ultrastructural changes to the cortical bone of the iliac crest after 5 years of treatment with risedronate. Despite the resolution advantages of in vitro imaging studies, the invasiveness of the procedure, inherent variability in specimen collection [28, 29], and the limited correlation to bone quality at clinically relevant sites for fragility fracture (proximal femur, lumbar spine, distal radius) [35] are drawbacks to the clinical application of micro-CT.

MDCT

MDCT is a clinical CT technique available in most diagnostic imaging departments, using scanners from a number of manufacturers (Table 1). Therefore, a dedicated scanner is not required. The spatial resolution of this technique is limited, with an in-plane resolution ranging from 150 to 300 μm and a minimum slice thickness of around 300 μm. These spatial resolutions are above trabecular dimensions, and imaging of individual trabeculae is subject to considerable partial-volume effects. However, given the larger size of intertrabecular spaces, trabecular bone parameters obtained with this technique are observed to correlate moderately well with those determined by contact radiography and micro-CT of bone specimens (r = 0.53–0.70) [71, 91], as well as micro-CT (r = 0.44–0.99) [43, 44, 69, 70, 85, 118]. The advantage of the MDCT technique is that central regions of the skeleton critically relevant to osteoporosis and fracture risk assessment, such as the spine [70, 73, 85] and proximal femur [5, 43], can be visualized. However, to achieve adequate spatial resolution and image quality, the required radiation exposure is substantial, offsetting the technique’s applicability in clinical, routine, and scientific studies (Table 1). High-resolution CT protocols are typically associated with an effective dose of approximately 3 mSv (1.5 years of natural background radiation), several orders of magnitude greater than standard DXA or HR-pQCT (4–13 μSv) and an order of magnitude higher than standard two-dimensional QCT (0.06–0.3 mSv) [41].

While early MDCT systems used for bone structure imaging typically consisted of four to 16 rows of detector elements [73], modern systems have 64 to 320 rows [69]. Most recently, high-resolution areal CCD detectors have been combined with a standard clinical CT gantry to provide substantial improvements in scan time and resolution [124]. In each case, the tomographic acquisition is performed with the subject lying supine on the scan table within the gantry. The xray source and detector ensemble continuously rotate about the field of view while the gantry translates along the rotational axis, effectively producing a helical series of projections [78]. Simultaneous calibration of Hounsfield units to mineral density is typically accomplished by placement of a solid hydroxyapatite phantom below the patient [49].

The analysis of trabecular microarchitecture from MDCT and flat-panel volumetric CT image data primarily involves the application of traditional histomorphometry [5, 17, 43, 70, 118], where BV/TV, Tb.N, Tb.Th, and Tb.Sp are calculated two-dimensionally using plate model assumptions [116]. In contrast, Ito et al. [73] used direct three-dimensional measures of trabecular dimensions, connectivity, and SMI in a MDCT study of the lumbar spine, finding a strong correlation (r = 0.98) between BV/TV measured by micro-CT and MDCT. Other specialized measures of trabecular dimensions using the fuzzy distance transform (which does not require a threshold binarization process) have been proposed by Krebs et al. [85].

To date, in vivo human studies using MDCT to assess bone structure are limited due to radiation dose concerns. Ito et al. [73] demonstrated SMI and BV/TV measured from MDCT images of the lumbar spine provided superior fracture discrimination to areal BMD by DXA. Graeff et al. [58] showed teriparatide treatment effects are better monitored by architectural parameters of the spine obtained through MDCT than by BMD (Fig. 3). In a cross-sectional cohort study of adolescent girls with and without anorexia nervosa, Bredella et al. [17] observed diminished trabecular microarchitecture at the distal radius in subjects with anorexia nervosa compared to controls, despite no differences in lumbar areal BMD. In a companion study, Lawson et al. [89] observed the abnormal trabecular microarchitecture in these patients are predicted by IGF-1, leptin, and androgen levels, with positive correlations (r = 0.32–0.72) to BV/TV, Tb.Th, and Tb.N.

Fig. 3A–D.

In vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.

HR-pQCT

A dedicated extremity imaging system designed for trabecular-scale imaging is currently available from a single manufacturer (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland). This device has the advantage of a higher signal-to-noise ratio and spatial resolution (nominal isotropic voxel dimension of 82 μm) when compared to MDCT. Furthermore, the radiation dose is lower when compared to whole-body CT and does not involve critical, radiosensitive organs in skeletally mature adults. There are several disadvantages to this technology. It is limited to peripheral skeletal sites and provides no direct insight into bone quality in the lumbar spine or proximal femur, common sites for osteoporotic fragility fractures. Additionally, there are currently a limited number of devices installed globally, which are primarily located at major research institutions with few available in clinical radiology departments.

In HR-pQCT, a standard protocol recommended by the manufacturer is utilized for most studies [14, 82]. The patient’s forearm or ankle is immobilized in a carbon fiber cast fixed within the gantry of the scanner. A single scout projection image of the distal radius or tibia is acquired to define the tomographic scan region. This scout image is acquired at an AP orientation at the wrist and at an oblique (45°) anterolateral-posteromedial orientation at the ankle. This tomographic region spans 9.02 mm in length (110 slices) and is localized to a fixed offset proximal from either the radial or tibial midjoint line and extends proximally. The offset is 9.5 mm in the radius and 22.5 mm in the tibia (Fig. 4). This method does not account for differences in bone length and may be a confounding source of variability in cross-sectional studies [16]. In the radius, the default axial scan location partially includes the most common site for fracture and location, where the bone microstructure is most strongly correlated to experimental strength of the forearm under a simulated falling load [106].

Fig. 4A–B.

Scout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.

There are several different protocol modifications for developmental studies in children and adolescents [26, 84] to account for patient size and age-related changes in bone length and to avoid radiation exposure to the epiphyseal growth plate and inclusion of provisional mineralized tissues from this region. In a cross-sectional study of age- and gender-related differences in the microarchitecture of the distal forearm of adolescents, Kirmani et al. [84] used a fixed offset (1 mm) with respect to the proximal extent of the distal epiphyseal growth plate of the radius. In contrast, Burrows et al. [26] selected a region offset of 8% of the total tibial length proximal to the tibial endplate. While there are a number of studies underway investigating other scan locations in adults, including more proximal sites dominated by cortical bone, the internal configuration of the XtremeCT gantry prohibits the positioning of true diaphyseal sites in the radius or tibia within the limited 15-cm longitudinal range of the source-detector ensemble.

The reconstructed images are analyzed using a standard protocol provided by the manufacturer. The operator chaperones a semiautomated contouring process to identify the periosteal boundary and segment the cortical and trabecular compartments [42]. The trabecular bone structure is extracted using an edge enhancement and threshold procedure [88]. While the importance of threshold selection for morphometric analysis is well described for other CT devices [61], few studies to date investigate threshold effects or propose new methods for HR-pQCT [23, 42]. The default compartmental and structural segmentation provides the basis for the subsequent densitometric, morphometric, and biomechanical analyses. For subjects with very thin or highly porous cortical bone, this segmentation procedure may fail to capture the cortical structure [42, 80]; therefore, more sophisticated autocontouring techniques that operate on the fine-structure segmentation have been proposed [18, 23]. For the XtremeCT, reproducibility of densitometric measures is very high (coefficient of variation < 1%), while biomechanical and morphometric measures typically have a coefficient of variation of 4% to 5% [14, 80, 82, 100, 105].

The linear attenuation values of the tomographic images are converted to hydroxyapatite mineral densities using a beam-hardening correction and phantom calibration procedure previously described for the ex vivo micro-CT system [21]. Based on this calibration, volumetric BMD is determined independently for cortical and trabecular bone compartments using the segmentation process described previously. HR-pQCT images are used to derive surrogate measures of areal BMD in the ultradistal radius [22]. This technique is associated with a high level of agreement (r² > 0.8) with multiple clinical DXA devices.

Unlike MDCT, which has a large slice thickness relative to the in-plane resolution, the high isotropic resolution of HR-pQCT (82 μm) permits direct three-dimensional assessment of intertrabecular distances. These measures were validated against micro-CT gold standards [23, 93, 97]. From the binary image of the extracted trabecular structure, three-dimensional distance transformation techniques are used to calculated trabecular number [65]. While the intertrabecular distances are large compared to the voxel dimension, the average trabecular thickness (100–150 μm) is only one to two voxels wide. Accordingly, direct measures of thickness and bone volume are complicated by considerable partial-volume effects. In the standard analysis protocol, BV/TV is derived from the trabecular volumetric BMD, assuming a fixed mineralization of 1200 mg hydroxyapatite per cm³ for compact bone (BV/TV = Tb.BMD/1200). Trabecular BMD is determined by calculating the mean mineral density of the full medullary compartment. From the direct measure of Tb.N and the densitometrically derived BV/TV, Tb.Th and Tb.Sp are derived using standard stereologic relations, assuming plate model geometry [87, 116].

There are several potential concerns with this approach. First, Sekhon et al. [134] documented substantial errors in the measurement of trabecular BMD related to biologically relevant variations in cortical thickness and the magnitude of trabecular BMD itself. This may be related to xray scatter effects and residual beam-hardening artifacts. These errors are primarily a concern for cross-sectional studies, when cortical thickness and trabecular BMD may span a broad range. It is less of a concern in longitudinal studies, where the percent of change as the primary end point, such as age-, disease-, and therapy-related changes in cortical thickness and trabecular BMD, is comparatively small. Second, the assumption of a fixed-matrix mineralization is inconsistent with the established action of many common antifracture therapeutics [10]. Changes in matrix mineral density is expected to cause an increase in BMD, irrespective of bone volume changes, and result in an overestimation of BV/TV and propagate error to the derivative measures of Tb.Th and Tb.Sp, confounding any actual therapy-related effect on trabecular bone volume and structure.

Several studies investigate other measures of bone microarchitecture and topology from HR-pQCT images, including connectivity, SMI, and anisotropy; however, there is mixed evidence of their reliability at in vivo resolutions [23, 93, 97, 136]. Recently, more sophisticated approaches to cortical bone segmentation have been proposed [18] that allow direct three-dimensional assessment of cortical thickness and quantification of cortical ultrastructure (Fig. 5), including intracortical porosity and canal diameter [24, 111]

Fig. 5A–D.

(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.

The ability of HR-pQCT to resolve the trabecular microarchitecture and a level of the cortical ultrastructure lends itself to calculating direct estimates of bone strength by voxel-based micro-FEA. For HR-pQCT scans of the distal radius and tibia, this is typically used to estimate strength under uniaxial compression, which approximates to the common loading condition for Colles’ fracture of the radius [106] and normal gaited loading for the tibia. This technique was validated against both higher-resolution models (based on micro-CT images) and empirical measures of strength [93, 99]. The application of micro-FEA is primarily performed assuming homogeneous material properties [15, 24, 40, 93, 99, 105], although it is also possible to produce models with material properties scaled according to mineralization [95, 99]. In addition to whole-bone mechanics, micro-FEA can be used to determine the relative load distribution between cortical and trabecular compartments [98] and estimate mechanical implications of specific structural features, such as the resolvable cortical ultrastructure [24].

For clinical investigations into longitudinal changes in HR-pQCT-derived measures of bone quality, it is critical baseline and followup scans be matched, since bone structure and geometry vary substantially along the axial direction [16]. Operator positioning for followup scans are aided by visual reference to the positioning of the baseline scan. Furthermore, postprocess image registration is performed to ensure comparable regions of interest are used for the image analysis. The manufacturer provides software that matches slices based on the periosteal cross-sectional area and limits the analyzed region to the slices common in baseline and followup [87]. Alternatively, MacNeil and Boyd [100] demonstrated three-dimensional image registration techniques can provide improved short- and medium-term reproducibility when compared to the default slice-matching approach. This approach may be more appropriate in longitudinal studies (long-term studies, anabolic therapy trials) where periosteal apposition would confound registration based strictly on cross-sectional area. As discussed earlier, the challenge in obtaining meaningful results in longitudinal studies in children or adolescents experiencing rapid growth is not trivial and requires careful consideration of standardized procedures for scan positioning and analysis [26, 101].

There is a growing body of literature featuring HR-pQCT assessment of bone quality. The first cross-sectional studies by Boutroy et al. [14] and Khosla et al. [82] reported gender-specific, age-related differences in trabecular bone microarchitecture. Several centers have observed age-related differences in micro-FEA estimates of bone strength in normative cross-sectional cohorts [24, 40, 96]. Furthermore, Burghardt et al. [24] and Macdonald et al. [96] demonstrated the ability of HR-pQCT to detect dramatic age-related differences in cortical porosity in females using new techniques for the analysis of cortical ultrastructure [19]. A microstructural basis for ethnicity-related differences in bone strength between East Asian and white women was reported in two studies [149, 150]. Sornay-Rendu et al. [137] suggested cortical and trabecular morphology provided additional fracture discrimination independent of areal BMD in osteopenic women. In the same cohort, Boutroy et al. [15] showed micro-FEA mechanical measures provided additional discriminatory power between osteopenic women with and without distal radius fractures. While the initial focus was predominantly related to fracture discrimination in postmenopausal osteopenia and osteoporosis [15, 103, 104, 137, 138, 140, 144, 145], a number of studies have utilized HR-pQCT to investigate developmental changes in bone quality and fracture risk [27, 34, 84], as well as secondary causes of bone loss [4, 20, 31, 62, 90].

Most recently, data from the first HR-pQCT single- and multicenter longitudinal trials were published. In a multicenter, head-to-head, randomized, placebo-controlled trial of denosumab (a RANKL inhibitor) and alendronate (a bisphosphonate), Seeman et al. [133] reported more pronounced antiresorptive efficacy with denosumab than alendronate. In particular, cortical thickness was preserved or improved at the radius and tibia with either treatment, while cortical bone loss progressed in the control group. Burghardt et al. [25] reported similar cortical bone changes and additionally showed a preservation of compressive bone strength by micro-FEA after 24 months of alendronate treatment. Longitudinal microarchitectural changes also occurred with strontium ranelate [125] and teriparatide [95].

Discussion

With the advent of new imaging equipment, image-processing methods, and mathematical modeling techniques, the importance of trabecular microarchitecture and cortical ultrastructure in the context of bone strength and fracture risk has received considerable attention. This review provided the essential background of high-resolution CT techniques now in use for clinical skeletal research and summarized the important clinical applications of this technology reported to date.

In focusing on the clinical applications of CT imaging, we did not directly cover a number of important applications of this technology in basic skeletal research. In particular, we did not review (1) numerous ex vivo cadaveric studies conducted to investigate structure-function relationships in cortical and trabecular bone; (2) preclinical imaging studies in animal models of disease, genetic intervention, and environmental effects on skeletal health; and (3) analysis of microarchitecture using texture analysis. Furthermore, it is important to recognize other imaging modalities, such as high-resolution projection radiography, MRI, and phase-contrast xray imaging, may also be used to assess bone microarchitecture. This review did not cover the substantial body of literature on these subjects.

Ex vivo micro-CT is the longest-standing technology used for nondestructive three-dimensional imaging of bone microstructure. Micro-CT scanners resolve the cortical and trabecular structure of human bone specimens and are used to evaluate aspects of bone quality in the context of skeletal disease and therapeutic interventions. However, their clinical utility is limited to small biopsy samples, typically taken from the iliac crest.

Translation of these techniques to in vivo imaging modalities, such as the MDCT and HR-pQCT, are subject to technical limitations related to image quality, radiation dose considerations, and subject motion. They provide images at resolutions approximately equal to Tb.Th (HR-pQCT) or intertrabecular distances (MDCT). Furthermore, while MDCT has the advantage of being able to image central skeletal sites, such as the spine and proximal femur, the images represent trabecular texture more than a true visualization of the individual trabecular structure. Similarly, HR-pQCT may not resolve the finest trabeculae or the complete scale of cortical ultrastructure features. Despite the challenges and limitations facing current in vivo CT imaging technologies, morphologic and biomechanical indices determined from these techniques correlate well with micro-CT [73, 97].

Finally, we demonstrated trabecular microarchitecture and cortical ultrastructure measured in vivo show age-, gender-, and race-dependent differences and provide improved fracture discrimination. Early longitudinal HR-pQCT observations suggest it is possible to detect structural changes induced by treatment. Ongoing studies and new results from therapeutic trials will provide a clearer indication as to whether trabecular microarchitecture and cortical ultrastructure measured using these in vivo methods will play a role in further understanding the affect of aging, disease, and interventions on skeletal health.

To date, there are no prospective fracture trials or large therapeutic trials used to draw conclusions regarding the role of in vivo assessment of trabecular microarchitecture and cortical ultrastructure in predicting bone strength and fracture status. These studies are warranted, and it would greatly enhance the field to establish a new set of diagnostic biomarkers that complement or improve upon areal BMD measured by DXA. The primary obstacles to achieving this goal are the substantial cost of large-scale human studies, the limited dissemination of the technology to clinical centers, the minimal standardization of image acquisition and image-processing protocols, and a lack of crosscalibration and quality control procedures (including standardized subject motion detection) for reliable data pooling in multicenter trials. Efforts to overcome the calibration issues, standardization, and multicenter comparative studies are needed to lay the groundwork for future large-scale multicenter trials and prospective studies.