Advances in Imaging Approaches to Fracture Risk Evaluation

Abstract

Fragility fractures are a growing problem worldwide, and current methods for diagnosing osteoporosis do not always identify individuals who require treatment to prevent a fracture and may misidentify those not a risk. Traditionally, fracture risk is assessed using dual-energy X-ray absorptiometry, which provides measurements of areal bone mineral density (BMD) at sites prone to fracture. Recent advances in imaging show promise in adding new information that could improve the prediction of fracture risk in the clinic. As reviewed herein, advances in quantitative computed tomography (QCT) predict hip and vertebral body strength; high resolution HR-peripheral QCT (HR-pQCT) and micro-magnetic resonance imaging (μMRI) assess the micro-architecture of trabecular bone; quantitative ultrasound (QUS) measures the modulus or tissue stiffness of cortical bone; and quantitative ultra-short echo time MRI methods quantify the concentrations of bound water and pore water in cortical bone, which reflect a variety of mechanical properties of bone. Each of these technologies provides unique characteristics of bone and may improve fracture risk diagnoses and reduce prevalence of fractures by helping to guide treatment decisions.

Introduction

Bone fractures are a widespread problem that affect over 75 million people in the world, with more than 2.3 million osteoporotic fractures per year globally (1, 2). Over a lifetime, the risk of a fracture is around 40% for women in developed countries (3). The costs associated with bone fractures were estimated to be $19 billion in 2005 in the United States alone, and are projected to increase by 50% by the year 2025 (4). In the EU, costs in 2010 were estimated to be €37 billion, and are expected to increase by 25% in 2025 (1). An increase in fracture risk occurs with aging for both women and men (4, 2). Fractures are a large problem with certain diseases and conditions, such as post-menopausal women and diabetes. Diabetes, in particular, has a rapidly increasing prevalence (5), leading to even higher costs and an increasing need for comprehensive clinical procedures to accurately measure and diagnose fracture risk.

The most common imaging parameter used to diagnose high fracture risk is low bone mineral density (BMD) assessed by dual energy X-ray absorptiometry (DXA) of the hip, spine, and distal radius. Examples of DXA images acquired in the radius and the hip are shown in Figure 1. DXA measures the transmission of X-ray beams through tissue at two different mean photon energies. The difference in dependence of X-ray attenuation on photon energy between bone mineral and soft tissues then allows for an estimate of BMD (6). Because DXA uses 2D projection images, the resulting BMD values are areal estimates, computed in units of mineral mass per image pixel area. In clinical practice, however, DXA BMD is typically evaluated as a T-score (tabulated over a standard region of interest), defined as an individual’s BMD relative to the standard deviation of BMD values of a young healthy population of the same ethnicity and sex (7). The World Health Organization has defined osteoporosis as having a T-score lower than −2.5 or having a previous fragility fracture, and osteopenia is defined as having a T-score between −1 and −2.5.

Figure 1.

Representative DXA scans acquired in the forearm (left) and the hip (right).

DXA is a fast, inexpensive, and well-studied method that has very low radiation dose (5–20 μSv), but it also has many limitations. Areal BMD varies significantly based on anatomical structure, so the results are biased by bone size and orientation. Degenerative disc disease or aortic calcifications can lead to an increased apparent BMD and falsely lower apparent fracture risk (8, 9), while other imaging artifacts arising from excess soft tissue in obese patients or prosthetic implants in the background can also alter DXA results. In addition, DXA does not fully explain the increase in fracture risk with age (10) or diabetes (11). Moreover, in a study of nearly 150,000 post-menopausal women (50 to 104 yo), 82% of those that reported a fracture within one year had a baseline T-score greater than −2.5 (DXA at peripheral sites, namely heel, finger, or forearm) (12).

To overcome some of the limitations of DXA, it is now standard of care to consider additional risk factors in the diagnosis and treatment of osteoporosis. This is often done using algorithms that incorporate known risk factors, such as The World Health Organization’s Fracture Risk Algorithm (FRAX) tool (13). This online tool calculates the 10-year probability of a major osteoporotic fracture and of a hip fracture based on relevant risk factors (e.g., age, sex, history of fracture, smoking status, alcohol consumption, and various diseases associated with high fracture) with or without hip BMD. The FRAX model is widely used in the clinic and is continuing to be expanded to include more countries. However, FRAX does not include all ethnicities or diseases, for instance type-2 diabetes, and is only designed to help guide clinical decisions. Other algorithms, such as Garvan and QFracture, have also been introduced as an alternatve to FRAX. The Garvan algorithm (14) was developed in Australia, includes the probability of suffering a fracture within both 5 and 10 years, and uses the history and frequency of previous fractures and falls. However, it does not include other risk factors and has only been tested on Australian and Canadian populations. The QFracture method (15) was developed in the United Kingdom and includes more risk factors than FRAX, such as various diseases, history of falls, and a 5 point scale for history of smoking and alcohol use. However, it does not include previous fractures in the model, and is limited to studies in the UK. In addition to the risk factor models, the trabecular bone score (TBS) is a gray-level texture measure that is derived from experimental variograms of DXA images of the lumbar spine. TBS is an indirect index of trabecular architecture, and has shown promise in adding to the predictive power of DXA (16). While both TBS and risk factor algorithms are useful tools, the primary limitation of these measures is that they lack additional information about the composition of the bone itself (17, 18).

Changes in both cortical and trabecular bone alter bone strength. With aging, there can be a thinning of the cortices, due to endosteal resorption, that leads to an increase in fracture risk (19). Aging can also lead to deterioration of the trabecular architecture (e.g., fenestrations of the trabeculae lowering the connectivity), thereby weakening the bone (20). Because bone loss usually begins in trabecular bone, clinicians are often interested in looking at trabecular bone measures to detect early changes in bone quality. Some more recent imaging methods have aimed to look at properties beyond areal BMD from DXA (21, 22).

Regardless of type, the bone tissue is comprised of three principal components: i) mineral (primarily crystals of calcium phosphate with carbonate and hydroxyl substitutions), ii) organic matrix (primarily type 1 collagen, non-collagenous proteins, and lipids), and iii) water (existing in porous spaces and bound to the matrix). The mineral component imparts strength and stiffness, and is the component of bone to which DXA is sensitive. However, the mineral component of bone alone is brittle; the plasticity or ductility of the bone comes from the hydrated organic matrix. During plastic deformation (i.e., post-yield strain), energy is dissipated until the bone fractures(19, 23). Along with bone structure, both the strength and plasticity of the bone tissue contribute to fracture resistance. Though increases in fracture risk are usually attributed to a decrease in BMD, changes in collagen organization or condition also affect fracture risk. For example, as a person ages, the collagen integrity of their bones decreases which results in increased brittleness of the bone (24, 25), leading to a significant increase in fracture risk. A major challenge in bone imaging is finding useful surrogates that are sensitive to bone brittleness.

This paper reviews additional imaging techniques that probe properties of bone that have the potential to help better diagnose fracture risk in clinical settings. The methods discussed are quantitative computed tomography (QCT) including high resolution peripheral QCT (HR-pQCT), quantitative ultrasound (QUS), micro magnetic resonance imaging (μMRI), and other quantitative MRI methods that provide information about the composition of the tissue. QCT methods provide 3D bone structure and volumetric BMD, which in-turn can also support the use of numerical methods to predict bone strength. High resolution HR-peripheral QCT (HR-pQCT) and micro-magnetic resonance imaging (μMRI) both assess the micro-architecture of trabecular bone. HR-pQCT also provides volumetric BMD, but at the cost of radiation exposure; μMRI has no ionizing radiation, allowing for repeated measurements, but has lower resolution and does not report BMD. Both HR-pQCT and μMRI have also been used in combination with μFEA to help improve bone strength predictions. QUS estimates of bone quality based on the ultrasound wave characteristics through bone tissue. Quantitative MRI methods can assess 3D bone structure, bone marrow fat content, and cortical bone water compartments including bound water and pore water components.

QCT and HR-pQCT

Quantitative computed tomography (QCT) uses conventional CT imaging applied in the lumbar vertebrae and proximal femur, concurrently with phantoms with known volumetric BMD values to convert image contrast into quantitative measures of volumetric BMD (mineral mass per image voxel volume) (26, 27, 28). More recently, opportunistic CT evaluations have been used to determine fracture risk (29, 30), in which CT scans acquired for reasons unrelated to osteoporosis are evaluated for low volumetric BMD in the spine or proximal femur. However, without a phantom, the scan is not quantitative, so this approach is only possible if the CT scanner is stable and calibrated regularly to ensure correct scaling of the intensity of the CT scans to Hounsfield units, which can be converted to BMD (29, 31, 30).

As a three dimensional measurement, QCT resolves whole bone structure and shape, but does not provide the resolution necessary to resolve trabeculae. Nonetheless, QCT can distinguish high volumetric BMD in cortical bone from low mean volumetric BMD in trabecular bone, and such measures of mean trabecular BMD in the proximal femur have been shown to discriminate between patients with and without hip fractures (32, 33, 34). QCT-derived volumetric BMD has also been shown to discriminate between patients with vertebral fractures and those with no fractures (35, 36, 37). In addition to volumetric BMD, measures such as cross sectional moment of inertia and cortical bone thickness can be determined. These measures have been shown to report on whole bone mechanical properties, as determined from tests on cadaveric bones (38, 33, 9, 34).

QCT-based finite element analysis (FEA) extends interpretations of QCT to incorporate whole bone structure (Figure 2). This computational approach for bone strength estimation has been reviewed in recent years (39, 40, 41) but a brief summary is given here. The voxels from a CT scan are used to define the finite elements that facilitate the numerical solution to the partial differential equations relating stress to the forces acting on bone (maintaining equilibrium) and relating strain to displacement (deformation of the bone). Each element is assigned material properties that define the stress-strain relationship or constitutive behavior of the bone tissue. These properties are typically based on empirical relationships between the apparent BMD and such elastic properties as modulus and yield strength. Boundary conditions simulating the forces experienced by a given bone in physiological conditions are applied, and a 3D mesh of coupled equations are solved numerically to estimate the stress and strain distribution in every element. This computation can be done using either linear or nonlinear models. Unlike linear models, nonlinear models allow plastic deformation to occur, but can require substantial computational time and memory to solve. A bone is considered to fail when it meets a certain fracture criteria such as exceeding an overall strain limit, or when a large enough volume of voxels exceeds an ultimate stress limit. Ex vivo studies have shown strong correlations between FEA predicted strength and experimentally measured strength of the hip (42) and of the vertebra (43). Cross-sectional studies of QCT-based FEA have shown the ability to discriminate between fracture and non-fracture cases (44, 45, 46), and may be able to predict fracture better than BMD alone (43).

Figure 2.

Assessment of incident spine and hip fractures in women and men using finite element analysis of CT scans: Sectioned views of finite element models of a vertebral body (left) and a proximal femur (right) showing the distribution of elastic modulus. Applied loads, which simulate axial compression for the spine and an unprotected sideways fall for the hip, are shown schematically, applied through layers of bone cement (white elements) to distribute the load over the bone surface. Reproduced from (46).

While QCT reports on the same mineral component of bone as DXA, it measures volumetric BMD, can be applied to both cortical and trabecular bone, and predicts whole bone strength. As a volumetric measure, QCT does not depend on the relative soft tissue attenuation, so it does not have the same issues as DXA with high attenuation in obese patients, or contamination from degenerative changes which can falsely increase areal BMD. QCT also allows for assessment of structural properties about the cortical bone size and shape.

Unlike QCT, high-resolution, peripheral QCT (HR-pQCT) can assess trabecular architecture and cortical porosity while also measuring volumetric BMD. HR-pQCT uses a dedicated imaging system for extremities at to achieve high resolution (80 μm isotropic) images of the distal tibia and/or distal radius. These systems provide both volumetric BMD and direct visualization of trabecular and cortical bone architecture, and an example of typical images in the distal radius can be seen in Figure 3. Cortical bone can be analyzed to assess parameters such as cortical thickness (Ct.Th), cortical porosity (Ct.Po) and cortical pore volume (Ct.Po.V). In trabecular bone, standard analysis includes quantifying structural properties of trabecular bone, such as bone volume fraction (BV/TV), which is derived from trabecular BMD (Tb.BMD), average number of trabeculae (Tb.N), average trabecular thickness (Tb.Th), and average trabecular separation (Tb.Sp) (47). Other trabecular properties have been explored as well, such as the connectivity (48), anisotropy (49), and structural model index (50), or individual trabecula segmentation (ITS) (51). The ITS software assesses contributions of plate and rod like structures in trabecular bone, such as ratio of rod to plate elements, orientation, plate and rod volume fractions and densities, and average size of plates and rods.

Figure 3.

Segmentation of forearm cortical and trabecular bone in HR-pQCT images. Reproduced in part from (164).

In comparison to DXA, HR-pQCT offers advantages in detecting early changes in trabecular bone in longitudinal studies in children (52) and transplant patients (53). Other examples of the utility of HR-pQCT include measurement of significant differences in trabecular architecture between post-menopausal women with and without history of fragility fractures (20, 54, 55), as well as differences in cortical porosity (56). Cortical pore volume (Ct.Po.V) was also found to be significantly higher in the tibia and radius of diabetic subjects (57). Following a period of disuse, it was found that detectable changes of the distal tibia microstructural properties occur, and recover when returning to normal weight bearing activity (58).

The isotropic voxels in a HR-pQCT image can be directly converted to elements for FEA, termed μFEA. Whereas the mesh of the bone is homogenized in QCT-FEA, the mesh in μFEA resolves bone from marrow spaces or pores. Linear μFEA can predict mechanical properties of the bone, such as the stiffness and strength (42, 59, 43), and have been shown to report on age and sex related differences (20, 60), to discriminate between fracture and non-fracture cases (61, 62, 63, 64), and to assess the severity of fractures (65, 66). μFEA has also shown potential for quantifying bone healing following a fracture (67, 68). Non-linear μFEA models have also been used to estimate post-yield parameters such as toughness (69). Because the finite element models are generated from axial scans of select regions of distal sites, the boundary conditions in μFEA do not necessarily simulate the loads that occur during a fall.

While QCT and HR-pQCT both have the ability to gain information about fracture risk, these methods have some limitations. QCT is more expensive than DXA, and central QCT has a high radiation dose, so measurements cannot be repeated frequently. Therefore, QCT is only advised in specific cases such as with obese subjects or advanced degenerative disease, and for monitoring metabolic changes in trabecular bone. HR-pQCT has a low radiation dose because radiosensitive organs are not close to the area of the scan (70), but is limited to the extremities, and is sensitive to signal contamination from motion or beam hardening. Also, like DXA, neither QCT nor HR-pQCT are sensitive to changes in the collagen matrix of bone. Both QCT and HR-pQCT are useful for finding more specific information about bone quality than DXA, but with the extra cost and radiation dose, they have not replaced DXA as a screening tool.

Quantitative Ultrasound

Unlike CT methods, quantitative ultrasound (QUS) is a low-cost method that is widely available. QUS measures both velocity and amplitude properties of ultrasound waves through bone tissue (71). The velocity of the measured waves, speed of sound (SoS), and broadband ultrasound attenuation (BUA) are the most commonly used measures to assess bone tissue, as well as values calculated from a combination of these two, the stiffness index (SI) and the quantitative ultrasound index (QUI) (72). These quantitative measures have been shown to reflect elastic modulus and compressive strength of the bone tissue (73). QUS is most often performed in the calcaneus of the heel, which is of particular interest because it is comprised primarily of trabecular bone, where bone loss usually begins, and easily accessible for a QUS measurement.

While calcaneous QUS has shown moderate correlations with BMD from DXA (71, 74), there has been a growing amount of research showing that QUS measures are an independent risk factor of fracture (75, 76, 77, 78). Additionally, QUS measures have shown differences between subjects with fractures and subjects without fractures in many studies (79, 80, 75, 77, 78). It has also shown some association with proximal hip fractures (81), and tends to be a better predictor of hip fractures than spine fractures (82). QUS has similar sensitivity as DXA, but lower specificity, and has been approved to be used to identify those in a low risk population where no further screening may be necessary (83, 82).

QUS is portable, inexpensive, has no ionizing radiation, and can be used in low resource settings where DXA is not available. While the amount of evidence for QUS as an independent marker of fracture risk is increasing, it is still not well standardized in the clinic, especially compared to DXA (80, 82). QUS results are also dependent on the devices, operators, anatomical locations, and positioning of the bone and ultrasound transducer. Clinically, it is not currently a recommended method for diagnosing osteoporosis or guiding treatment decisions (84, 80) though more recently it has been suggested as a screening tool for fracture risk assessment (83, 82). While it offers a radiation-free method of assessing bone, the efficacy of the methods need to be improved and further studied before it is a clinically viable tool (80, 85).

μMRI

Micro-MRI, or μMRI, evaluates both cortical and trabecular bone properties, such as cortical thickness and trabecular bone microarchitecture (86, 87, 88, 89), and performs similarly to HR-pQCT (90). Several pulse sequences have been used for high resolution structural imaging of bone, including spoiled gradient echo (91, 92), balanced steady state free precession (b-SSFP) (93, 94), and fast large spin echo (FLASE) (95, 96, 97). These pulse sequences all allow images to be acquired at a relatively high resolution (100–200μm in plane resolution). Unlike gradient echo or b-SSFP, spin echo methods such as FLASE are less sensitive to off resonance effects that can cause distortions in the trabecular architecture. However, spin echo sequences have a longer minimum acquisition time because the TR required is much longer (94). μMRI of bone is usually acquired at extremities such as the distal tibia or distal radius, but it has also been applied at the proximal femur (98, 99, 100), as shown in Figure 4.

Figure 4.

Representative coronal high-spatial-resolution 3-T MR images of proximal femur microarchitecture in a subject with osteoporotic fracture (left panel) and a control subject (right panel). Trabeculae are hypointense linear foci. There is deterioration in trabecular microarchitecture in the fracture subject compared with the control subject. Reproduced from (165).

Similar structural and architectural measurements as HR-pQCT can be derived from μMRI images (90). These measures of bone derived from μMRI have been shown to report on fracture risk. For example, many studies assessing achitectural properties in the distal radius have shown differences between fracture and non-fracture cases in the vertebrae (101, 102, 54), and other studies have shown changes in properties in response to drug treatment (103, 92, 104, 105). Recently, μMRI has been used with machine learning to improve the accuracy of discriminating between fracture and non-fracture cases (106). μMRI can also be used with μFEA to predict mechanical properties from structure (107, 108). For example, in renal transplant patients, μFEA showed significantly lower mechanical properties after transplantation (109), even when structural properties did not change significantly.

While μMRI probes similar characteristics of bone as HR-pQCT, μMRI has an advantage because it does not involve ionizing radiation, which enables repeated scans over a short time span and application to the proximal femur (98, 110, 111, 100). However, μMRI does not report on BMD, has lower spatial resolution, and is more expensive than HR-pQCT. Also, current μMRI methods are sensitive to magnetic susceptibility artifacts which can change the apparent size of trabeculae(94, 112), and motion artifacts due to the long scan times (113, 114). Overall, use of μMRI in the clinic is limited by resources and cost, but it has the potential to greatly add to the ability to predict fracture risk.

Other Quantitative MRI Methods

The imaging measurements of bone quality discussed thus far are only sensitive to the mineral composition of bone. While imaging the mineral component allows for measures of BMD as well as many structural changes that relate to bone strength, other components of bone such as the collagen content and the fat content can report on fracture resistance. Quantitative MR measures are sensitive to water and fat in the bone, including bone marrow fat and collagen-bound water in cortical bone. It has been shown that there is a relationship between marrow fat content and fracture risk (115), because the adiposity of marrow is related to bone metabolism (116). Vertebral marrow fat fraction increases with age (117), as does the portions of saturated, monosaturated, and polyunsaturated fat (118). Generally, greater proportions of saturated marrow lipids are associated with increased fracture risk (119) and with lower DXA BMD (120).

MR spectroscopy (MRS) measures signal from a volume of tissue and then uses spectral analysis to determine the molecular sources of the signal—the signal amplitude at particular frequencies are related to the volume of a particular chemical components in the tissue. In particular, MRS of bone marrow can measure both the volume fraction of fat and the relative amounts of unsaturated and saturated fats (121), and has shown good reproducibility in the vertebrae (122, 123). Several MRS studies have shown a significant association between marrow fat fraction and prevalence of fractures (124, 118, 123, 125). MRS has also found altered bone marrow fat composition in type-2 diabetes cases, (126, 127, 125), where DXA tends to give higher BMD levels despite the increase in fracture risk. MRS has the advantage of being widely available on clinical MR systems, so it is relatively straightforward to implement.

In addition to the water and fat in marrow, MRI can probe both the water bound to the organic matrix (bound water) and the water residing in the pore space (pore water) of cortical bone. MRI images of cortical bone water in the tibia are shown in Figure 5. In cortical bone, the proton signal from different components can be distinguished by transverse relaxation time constant (T₂ ) (128, 129, 130). The collagen protons make up the pool with the shortest relaxation times (T₂ < 100 μs); the water bound to the collagen matrix of cortical bone has a relatively short T₂, 100 to 1000 μs; and the water in the pore spaces of cortical bone span wide range of T₂ values from 1 ms to 1 s, reflecting the wide range of pore sizes (131).

Figure 5.

Representative transverse MR images of tibial midshaft, with and without softtissue signal suppression. A, Gradient-echo (GRE) image. B, Radial ultrashort echo-time images. C, Radial ultrashort echo-time images with soft-tissue suppression (suppr.). Circular structure is the reference sample with T2 at approximately 300 μsec which, similar to bone, is visible only on radial MR images. Reproduced in part from (139).

In conventional MRI scans, the echo time (TE) is longer than the T₂ of most of the bone water signal, so little or no bone signal can be seen in the image. However, imaging sequences such as ultra-short echo time (UTE) (132), water and fat suppressed projection imaging (WASPI) (133), zero echo time (ZTE) (134), and SWeep Imaging with Fourier Transform (SWIFT) (135) can reach effective echo times of 100 μs or less, permitting measurement of the bound and pore water signals, in particular. Most in vivo studies to date have used UTE (136, 137, 138, 139, 140, 141, 142, 143, 144), which is relatively easy to implement on clinical MRI systems. UTE imaging achieves a short TE by starting signal acquisition immediately following the radio-frequency excitation (and a short delay for switching the electronics from transmit to receive mode). Data are acquired while the read-out gradient ramps from zero, which means that k-space is sampled radially from the center and no delay is necessary for a phase-encoding or read-preparation gradient pulses. This also means that there is no time for a slice-select refocussing gradient pulse, so UTE requires either a 3D acquisition using non-selective excitation pulses or special slice-selective pulses that do not require gradient refocussing (discussed below). Naturally, UTE is not amenable to spin echo acquisition, so the signal amplitude as a function of TE depends on $T_2^{\ast }$ not T₂.

Early UTE studies of cortical bone characterized total signal, but an NMR relaxometry study of cadaver bone samples demonstrated that the bound and pore water signal amplitudes change in opposite direction with change in bone mechanical properties (145, 146). Thus, UTE MRI methods that best predict fracture risk should robustly distinguish and quantitatively measure bound- and pore-water concentrations in cortical bone. An overview of methods developed for this purpose was recently presented in Seifert et al (147) and is briefly highlighted below.

One approach is to distinguish the bound and pore water signals by $T_2^{\ast }$ rather than T₂ weighting. In the simplest form, two UTE images are acquired, one at the shortest achievable TE and one with a sufficiently long TE such that only pore water signal remains. A ratio of these two image intensities has been called the porosity index and corresponds well with the pore water fraction as well as μCT measures of porosity (144). A similar measure of relative amplitudes of bound and pore water through non-linear regression of intensities from several images with different effective TEs has also been implemented (148, 137, 149, 150) and has shown good correlations with mechanical properties of bone (141). These $T_2^{\ast }$ based approaches are essentially equivalent to early non-localized studies of bone samples at low magnetic field (151); however, line broadening will tend to result in an overlap of bound and pore water signal $T_2^{\ast }$ values making these approaches less effective at higher magnetic fields (152, 153, 154).

Alternatively, bound- and pore-water UTE MRI methods that discriminate bound and pore water signals based on T₂ (not $T_2^{\ast }$) using adiabatic RF pulses was proposed by Horch et al. (152) and later translated (155), validated (143), and shown to report on various mechanical properties of bone, including toughness (156). This approach is an extension of earlier UTE methods that used adiabatic RF pulses to suppress long T₂ signals from tissues such as muscle and bone marrow to enhance visibility of tissues with short T₂ signals, such as bone (157, 140, 158). Specifically, the adiabatic inversion recovery (AIR) method, for bound water imaging, uses an adiabatic full passage pulse followed by an appropriate delay to invert and null pore water magnetization while the bound water magnetization recovers from zero to near full amplitude. Similarly, two consecutive adiabatic full passage pulses will drive the bound water magnetization to zero while rotating pore water magnetization through 360, leaving it essentially unaffected. This approach is referred to as the Double Adiabatic Full Passage (DAFP) method and is used to image pore water. Both of these methods, similar to some previous and subsequent UTE studies of bone, use a reference marker with a known proton concentration in the imaging field of view in order to convert image intensity into a measure in absolute units of concentration.(139, 140, 159, 155). Representative bound and pore water concentration maps from in vivo acquisitions in the tibia and radius are shown in Figure 6.

Figure 6.

Axial MR images obtained in vivo in the lower leg (left) and wrist (right) in a healthy subject; images are conventional UTE images overlaid with pore (top row) and bound (bottom row) water maps in the tibia and the radius. Reproduced in part from (143). 45

While these quantitative UTE MRI methods are advantageous because they are sensitive to different components of bone than the other imaging methods discussed in this article, they have limited resolution and have only been applied in vivo in the radius and tibia. Also, MRI is expensive and can result in relatively long scan times compared to X-ray based methods. Using 2D UTE rather than 3D UTE can dramatically reduce scan times, but 2D UTE is known to be highly sensitive to the performance of the magnetic field gradients (136, 160, 161). Recent technical developments in gradient waveform optimization (162) applied to 2D UTE have demonstrated the potential for quantitative bound and pore water imaging in the tibia in <1 min of scan time (163). This 2D UTE method has the potential to be applied in the femoral neck, where most traumatic fractures occur, though cortical bone in the femoral neck is thin, and these MRI methods need more development to be applicable in such areas. Beyond just the development of 2D UTE methods, continued developments and evaluations in clinical subjects are necessary to to determine the ultimate utility of quantitative UTE MRI.

Conclusion

A summary of the imaging methods discussed in this article are shown in Table 1. These emerging imaging methods have the potential to provide better fracture risk assessment than current clinical techniques. HR-pQCT and μMRI can help by providing more information on bone structure, particularly in trabecular bone microarchitecture. QUS offers information about the quality of bone at low cost. MRI methods for quantifying fat could also help to independently characterize fracture risk, especially in diabetic patients. Methods for imaging cortical bone using MRI can probe new information about the material properties of the bone, since they are sensitive to the bound and pore water components rather than the mineral component. To date, the evidence for the efficacy of the imaging methods to accurately assess fracture risk come from case-control studies. Moving forward, longitudinal, prospective studies are necessary to determine at what thresholds the measurements from advanced imaging techniques indicate when a patient requires fracture prevention therapy. Future work should refine these imaging methods to further enhance their sensitivity to fracture resistance of an indivdidual’s bone.. The non-ionizing radiation methods are also well suited for longitudinal studies, which could help monitor disease progression over time, thereby assisting clinicians in deciding when an intervention is needed or when a drug treatment can be stopped. Imaging methods could also be used with personalized drug therapies to better prevent fractures or evaluate disease state. Newer imaging methods may also improve monitoring of the fracture healing process. In conclusion, developing new imaging methods to evaluate bone fracture risk could yield better and safer methods for treatment planning in cases of osteoporosis, diabetes, and other diseases associated with increased fracture risk, to ultimately reduce fragility fractures in patients.

Table 1.

Summary of imaging modalities with respect to advantages, disadvantages, and characteristics assessed

Imaging Method	Ionizing Radiation	Cost/ Time	Structural Characteristics	Material Characteristics	Associated Tools
DXA	low	low	—	aBMD	TBS, risk factor models+
QCT	high	high	macro-structure	vBMD	FEA
HR-pQCT	low*	high	macro-structure, porosity, trabecular architecture	vBMD	μFEA
QUS	none	low	—	SoS, BUA	QUI, SI
μMRI	none	high	macro-structure, trabecular architecture	—	μFEA
Marrow MRI	none	high	—	fat fractions	—
Bound/Pore Water MRI	none	high	macro-structure	bound water, pore water	—

^*limited to extremities to maintain low dose

⁺These tools (e.g. FRAX, Garvan, QFracture) have only been used with DXA, but could potentially be applied to other imaging modalities.

Abbreviations

DXA

dual energy x-ray absorptiometry

BMD

bone mineral density

FRAX

fracture risk algorithm

HR-pQCT

high resolution peripheral quantitative computed tomography

QUS

quantitative ultrasound

SoS

speed of sound

BUA

broadband ultrasound attenuation

SI

stiffness index

QUI

quantitative ultrasound index

NMR

nuclear magnetic resonance

MRI

magnetic resonance imaging

MRS

magnetic resonance spectroscopy

UTE

ultrashort echo time

DAFP

double adiabatic full passage

AIR

adiabatic inversion recovery