Cortical bone MRI, especially for the estimation of cortical thickness for the proximal femur, may ensure greater accuracy since the image plane can be oriented perpendicular to the neck [6]. Using advanced MRI methods with ultra short echo times (UTE), the bone water content in the microscopic pores of the haversian and the lacunocanalicular systems of cortical bone can be quantified. They contain approximately 20 % water by volume [7, 8]. A smaller water fraction is also bound to collagen and the matrix substrate and imbedded in the crystal structure of the mineral [9]. These micro pores maybe difficult to visualize due to their size compared to the spatial resolution of the images, but the quantification of bone water using MRI could potentially provide a surrogate measure of bone porosity without resolving these individual small pores.

Image analysis of trabecular bone images in MRI involves the selection of initial separate regions of interest consisting of trabecular bone and cortical bone, followed by corrections of spatial signal homogeneity corrections, if needed, and image separation into a bone and a marrow phase. Following this, a number of characteristics of the trabecular and cortical bone structures may be derived. If the MRI images are generated from longitudinal studies over time, the subsequent images may require serial image registration [10]. Image analysis techniques to segment both the inner and outer cortical boundary from their surroundings are commonly semi-automatic. A software algorithm featuring a deformable contour (snake) to conform to the strongest gradient edges in the neighborhood of the manually placed cortical bone region of interest has been used [11]. Another method used for image analysis is the distance transform method [12] where the measurements of cortical thickness are further improved by fitting a sphere into the segmented cortical shell. Thickness measurements from MR data have been compared with HR-pQCT, and while significant correlations were found, the values for the MRI measures were higher [13].

For the assessment of structural information of trabecular bone, three classes of parameters can be defined including scale, topology, and orientation [14, 15]. Scale describes the amount of bone in a region of interest (ROI) and the thickness of the trabeculae or the spacing between the trabeculae. Investigating the plate- or rod-like structure of the network assesses the bone topology. And finally the anisotropic character of the structure defines its orientation. Early assessments of trabecular bone applying the principles of stereology are based on scale [16]. In MRI, trabecular thickness can be obtained from the mean intercept length (MIL) of parallel test lines across the ROI averaged over multiple angles [17, 18]. From MIL and BV/TV measurements, trabecular thickness (TbTh), trabecular spacing (TbSp), and trabecular number (TbN) can be obtained [19]. These measures are usually called “apparent” parameters because of the limited spatial in vivo resolution of MRI. These measurements are usually conducted on a slice-by-slice basis. Other approaches have been proposed such as 3D wavelets analysis [20], fuzzy distance transform [21] have also been proposed.

Describing bone structure through bone topology is useful for osteoporotic bone loss which results in a fenestration of trabecular plates and a conversion from plate to a more rod like bone structure [16, 22–24]. Methods for assessing trabecular bone connectivity and digital topological analysis (DTA) have also been suggested [25, 26]. The technique was successfully applied to trabecular bone topology [27, 28] and enhanced the prediction of mechanical properties and bone strength [29]. A more recent technique provides a complete assessment of scale, topology, and anisotropy using geodesic topological analysis (GTA) [30].

The developments over the past few years have made quantitative MRI of bone clinically practical [27, 31–41]. A substantial improvement in fracture discrimination by including structural information in addition to bone mineral density (BMD) has been well established [36–38]. The effect of salmon calcitonin on bone structure was investigated using MRI at the distal radius and calcaneous of 91 postmenopausal women over a period of 2 years [42]. The treatment group showed improved trabecular structure compared to the placebo group but no significant change in BMD was detected. Topological changes of the trabecular bone network after menopause and the protective effect of estradiol were reported by Wehrli and colleagues in 2008 [43]. The effect of testosterone replacement on trabecular architecture in hypogonadal men was investigated in the distal tibial metaphysis of 10 severely testosterone-deficient hypogonadal men [32]. A subvolume of each MR image was converted to a finite element model. No significant changes in estimated elastic moduli and morphological parameters were detected in the eugonadal group over 24 months but a significant increase in four estimated elastic moduli was found in hypogonadal men. These increases were accompanied by significant increases in trabecular plate thickness.

In a pediatric population of 40, 6–12-year-old subjects, high-resolution magnetic resonance images were collected immediately above the growth plate in the distal femur [44]. Measures of trabecular bone microarchitecture [i.e., apparent trabecular bone volume to total volume (appBV/TV), trabecular number (appTb.N), and trabecular separation (appTb.Sp)] showed strong correlations with the distance from the growth plate reflecting the spatial heterogeneity of trabecular bone (Fig. 13.2) [45]. Gender differences were not found in MRI-based measures of trabecular bone microarchitecture, consistent with the BMD measures by DXA at the distal femur [45]. MR derived measures were moderately to strongly related to aBMD and BMC. In a separate study, Modlesky et al observed children with cerebral palsy had a 30 % lower appBV/TV, a 21 % lower appTb.N, a 12 % lower appTb.Th and a 48 % higher appTb.Sp in the distal femur than controls (n = 10/group; p < 0.001) [46]. The short-term reliability of the trabecular bone microarchitecture measures was very good, with coefficients of variation ranging from 2.0 to 3.0 % in children with CP (n = 6) and 1.8–3.5 % in controls (n = 6).

In a study investigating the cortical thickness of 41 postmenopausal osteopenic women and 22 postmenopausal osteoporotic women with spine fractures, significant changes in the cortical thickness between the two groups were found underlying the importance of morphologic measurements of the cortical bone structure [11]. In 2009, images of the distal radius and the distal tibia of 49 postmenopausal osteopenic women (age 56 ± 3.7) were acquired with both HR-pQCT and MRI [47]. It was found that the amount of cortical porosity did not vary greatly between subjects but the type of cortical pore containing marrow versus not containing marrow, varied highly between subjects. Additionally, the number of cortical pores containing marrow did not depend on the amount of porosity and there was no relationship between cortical pore size and the presence of bone marrow. The data suggest that cortical pore spaces contain different components and that there may be more than one mechanism for the development of cortical porosity and more than one type of bone fluid present in cortical pores. However, this approach only captures relatively large cortical pores, which can be visualized within the resolution limits of MRI.

In a recent study measurements in vivo revealed that the bone water content was increased 65 % in the postmenopausal group compared to the premenopausal group [41]. Patients with renal osteodystrophy had 135 % higher bone water content than the premenopausal group whereas conventional BMD measurements showed an opposite behavior, with much smaller group differences.

In a study with non-Hispanic white adolescent females (N = 24; 18–19 years of age), two bone-specific retrospective physical activity loading tools were used, the Bone Loading History Questionnaire (BLHQ) and the Bone-Specific Physical Activity Questionnaire (BPAQ), along with a 7 day physical activity recall to assess energy expenditure. Hip BLHQ scores were correlated with mid-tibia cortical volume assessed by MRI (r = 0.43; p = 0.03) [48]. Adjusted hip and spine BLHQ scores were correlated with all mid-tibia cortical measures (r = 0.50–0.58; p < 0.05) and distal radius apparent trabecular number (r = 0.46–0.53; p < 0.05). BPAQ scores were correlated with all mid-tibia cortical (r = 0.41–0.51; p < 0.05). These data demonstrated that greater load-specific physical activity scores, but not energy expenditure, are indicative of greater mid-tibia cortical bone quality.

MR imaging is a noninvasive method for three-dimensional imaging of both trabecular and cortical bone, and can be used for assessing skeletal growth, metabolic diseases, and other bone-related diseases in a pediatric population. The added advantage of being nonionizing and the potential for characterizing muscle, bone marrow as well as body fat makes this an exciting tool for the pediatric musculo-skeletal system.

Finite element analysis of patient computed tomography (CT) scans has been in development in academia for over 25 years [49–55] and is now becoming available clinically for assessing fracture risk and monitoring treatment. This analysis technique combines medical image processing of CT scans, bone biomechanics, and the engineering finite element analysis technique [56] to provide a “virtual stress test ” [53] of a bone under a prescribed set of external forces. The primary outcome of the analysis is an estimate of the breaking strength, in units of force (i.e. Newtons) of a patient’s whole bone or portion thereof, e.g. a vertebral body, proximal femur, or distal radius or tibia.

When analyzing a CT scan with finite element analysis, measures of volumetric bone mineral density (BMD, in mg/cm³) can also be provided, as well as measures of DXA-equivalent areal BMD (in g/cm²) and the associated T-scores [53, 57–61]. By combining measures of both BMD and bone strength, the overall analysis provides a comprehensive noninvasive assessment of bone competence, and can be applied as an “add-on” analysis to previously taken clinical CT scans that are originally ordered for some other nonbone indication, such as a gastroenterology exam [58].

In a research setting, finite element analysis has been used in a number of clinical studies to longitudinally monitor changes in bone strength in response to various therapeutic treatments. By altering the finite element structural models in a controlled fashion, for example, by virtually removing an outer layer of bone or by averaging out the spatial distribution of bone mineral density within the boundaries of the bone—the analysis can also provide measures of bone strength associated with changes in just the cortical bone, the trabecular bone, or the external bone geometry, all of which can provide unique insight into changes related to growth, aging, disease, and treatment [62–64].

This review summarizes some of the key principles involved in finite element analysis of bones from CT scans of live people and its application to pediatric patients. For more detail on the overall approach, the reader is referred to several reviews [53, 54, 57, 65, 66].

In general, there are three different types of finite element models that can be generated from CT scans, depending on the spatial resolution of the CT scan that is used as input to the analysis. Starting at the highest spatial resolution, using images from a highly specialized “micro-CT” scanner as input, one can generate finite element models having a spatial resolution as low as 20 μm [67]. Such high-resolution models can only be generated for small specimens of cadaver bone and thus this type of finite element analysis is not applicable to clinical studies on live people (and will not be reviewed further in this chapter). However, one relevant result to the current discussion is that the models can predict strength very well, with R ² values of at least 0.85 at both the hip and spine [68, 69]. Since those models do not contain any features at a resolution of less than about 20 μm, variations in any patient-specific factors at that scale or below—for example, bone tissue material properties, collagen cross-linking, damage, or geometric features of lacunae or the remodeling space—do not appear to play any appreciable role in the overall whole-bone strength. That is, at least for these cadaver studies. Clearly, certain disease states or perhaps other factors may alter these lower-scale properties and affect overall bone strength. But it appears otherwise that typical patient-specific variations in these low-scale features may have little influence on overall whole-bone strength.

At a slightly lower resolution, models having a resolution on the order of 100 μm can be generated from “high-resolution peripheral quantitative CT” (HR-pQCT) scanners [65]. These are also specialized types of CT scanners but they are clinically available at some medical centers. While these types of finite element models can include patient-specific descriptions of the trabecular microarchitecture, cortical thickness, and even cortical vascular porosity, because of constraints on the size of a body part that can be scanned, the input scans can only be obtained for the extremities in live humans, most commonly at the distal radius and tibia. Because of the relatively low radiation exposure associated with HR-pQCT scans, these types of finite element analyses have now been used in a number of clinical research studies in pediatric research applications.

At the lowest level of resolution, one can generate finite element models from clinical CT scanners, having a spatial resolution on the order of 1 mm. These models typically cannot provide explicit descriptions of the trabecular microarchitecture, nor of the cortical porosity, and do not have sufficient resolution to accurately capture the cortices where they are thin (<0.5 mm), such as in the vertebral body and in portions of the proximal femur [70]. Even so, because these types of finite element models can be generated for any part of the body—particularly the hip and spine—and because they capture both the three-dimensional shape of a whole bone and its internal spatial distribution of bone mineral density, and because they have been in development for so long [51], they have found the most widespread use so far in research studies in adults and are the only types of finite element models currently approved by FDA for clinical diagnostic purposes. However, partly because of concerns about radiation exposure associated with clinical CT scanning of the hip or spine regions in growing children, these types of finite element analyses have not been utilized much in pediatric research applications.

In all three types of finite element analyses, the same general “voxel-conversion” technique is typically used to generate a patient-specific model of a whole bone, or portion thereof, from the CT scan. In this technique, the attenuation of the input CT scan is calibrated into units of bone mineral density and then each image voxel (i.e. 3D equivalent of a pixel) is converted directly into a box-shaped finite element, perhaps with some resampling if the desired size of the finite element differs from the image voxels. The images are then registered into a common coordinate system to enable virtual forces to be applied in a standardized fashion. Finally, mechanical properties of the bone tissue are assigned to the individual finite elements based on the calibrated BMD values of the individual voxels. Precisely how this last step is done depends on the spatial resolution of the CT scan and what assumptions are made about the mechanical behavior of the bone tissue, and what relation is used to map the calibrated BMD for each finite element into mechanical properties, all of which can differ at different anatomic sites and with different software implementations of the finite element analysis. The result is a “voxel-based” finite element model (Fig. 13.3), which can be generated in a highly automated fashion due to the direct approach of converting image voxels into finite elements. While more complex methods can be used to process the images and create the finite elements [71], this voxel-conversion approach is currently the most widely used approach for patient-specific analysis and is discussed further below for use with HR-pQCT and clinical CT scans.

HR-pQCT scans are typically acquired for the peripheral skeleton, for example, the distal radius and/or tibia. These images currently have a spatial resolution of about 80 μm. In the process of converting image voxels to finite elements, the bone is virtually separated from the marrow space, resulting in a finite element model that contains only bone elements. Thus, the microstructure and pore spaces are explicitly captured with this approach (Fig. 13.4), although the resolution is not ideal since individual trabeculae have a thickness on the order of the element size and thus substantial volume-averaging can occur at all bone-marrow interfaces. Despite this limitation, a number of cadaver studies have demonstrated that the technique can provide good correlations (R ² = 0.66–0.97) between its predicted strength of the distal radius during a simulated fall on the outstretched hand and the strength as measured by direct mechanical testing [55, 72–75], and these correlations are typically higher than those provided by any BMD measures [66].

Clinical validation studies show more mixed results, especially when compared to BMD as measured by DXA [66]. In a series of studies that assessed distal radius fractures in postmenopausal women, the odds ratios for finite element analysis and DXA-BMD (at the wrist) were both statistically significant but were similar to each other [74, 76, 77]. When assessing any type of osteoporotic (prevalent) fracture, odds ratios for finite element analysis and DXA-BMD (at the wrist) again were both statistically significant but were either similar to each other or trending (but not statistically significant) higher for finite element analysis [78–80]. Some results suggest that the finite element analysis performs slightly better when used at the distal tibia, and there may be advantages to developing predictive algorithms that combine together various different outcomes from the finite element analysis [74, 80]. No prospective incident-fracture outcome validation studies have yet been reported and thus this represents an important topic for future research. Due in part to the limited availability of HR-pQCT scanners, and since this analysis technique cannot be used to assess changes at the hip or spine in live humans, these types of finite element models have so far found relatively limited use in clinical research studies in adults for assessing treatment effects [81–87]. Even so, the technique does provide unique insight into biomechanical consequences of any treatment-related changes in bone microstructure and has been validated for assessing treatment-related changes in strength in a large animal model [88].

For clinical CT scans, central sites can be imaged and the sites of most interest for assessment of osteoporosis are the proximal femur and vertebral body, which can be rapidly scanned with minimal motion artifact using contemporary CT scanners. The typical resolution of the resulting images is about 1–2 mm. Thus, in these types of finite element analyses, the bone tissue and marrow are volume-averaged at that scale, so that the models do not include any explicit descriptions of the microstructure or porosity. Instead, the porosity of the bone is represented by intra-bone variations in the levels of bone mineral density throughout the bone, and any real presence of blood or marrow is accounted for by calibrating the attenuation values to units of attenuation-equivalent BMD. Despite these limitations, this type of finite element analysis has also been well validated in cadaver studies and clinical studies. For example, cadaver experiments from multiple research groups have consistently shown good predictions of experimentally measured femoral strength (R ² = 0.75–0.96) [50, 89–96] and vertebral strength (R ² = 0.75–0.96) [49, 97–101].

Clinical validation studies have consistently shown that prevalent vertebral fractures [51, 64, 102–104] and new (incident) vertebral [61, 97] and hip [60, 61, 105] fractures, in women and men, are highly associated with finite element-estimated strength and almost always more so than DXA-measured BMD. Prospective fracture-outcome studies have also demonstrated that finite element analysis-derived measures of vertebral and hip strength provide additional diagnostic information beyond BMD by identifying patients without osteoporosis who have “fragile bone strength,” that is, low levels of vertebral or hip strength that place them at a high a risk of a new vertebral or hip fracture as a patient with BMD-defined osteoporosis [58, 61]. Because clinical CT scanners are so widely available, and because of the keen interest in evaluating therapeutic-related changes in bone strength at the hip and spine in older adults, these types of finite element models have been quite widely used in clinical research studies in adults to assess longitudinal changes in bone strength in response to various osteoporosis therapeutic treatments [63, 103, 106–118].

Since the same CT scans used for finite element analysis can also be used to measure a DXA-equivalent BMD T-score at the hip, [53, 57–61], and a vertebral trabecular BMD at the spine [61, 119, 120], these types of quantitative analysis of CT scans provide measures of both BMD and bone strength, resulting in a more comprehensive clinical assessment of bone quality than one based on BMD alone.

In general, finite element analysis has been applied to a wide variety of pediatric applications, ranging from injury criteria for trauma and motor vehicle accidents, to surgical planning, to the study of etiological factors in various topics related to growth and development. Examples of patient-specific analyses (Table 13.1) as applied to the biomechanics of growth and development include studies on athletes, anorexia, genetics, hormonal influences, activity levels, and fracture etiology. Mostly, HR-pQCT is the scanning method of choice, although some patient-specific analyses have utilized X-rays (multiplanar) and clinical (quantitative) CT scans.

These finite element studies have provided unique insight into fracture etiology. For example, in a study of 198 boys and girls ages 8–15 [121], results suggested that fat mass may have a different adaptive influence on whole-bone strength at weight-bearing (e.g. distal tibia) compared to non-weight-bearing (e.g. distal radius) sites. This in turn leads to relatively weak bones at the distal radius in relation to body mass in obese children—which would explain why obese children are at higher risk of fall-related wrist fractures. In a study of anorexia involving 44 adolescent girls [122], distal radius bone strength from finite element analysis was significantly lower in girls with anorexia than in a group of control girls without anorexia, even after controlling for distal radius BMD as measured by DXA. The underlying mechanisms were due in part to alterations of both the cortical and trabecular microarchitecture, including a lower cortical area.

Given the relative novelty of patient-specific finite element analysis for pediatric applications, and the continuing improvement of CT scanning technology in terms of delivering higher quality images at lower radiation doses, it is expected that much new insight will be gained as patient-specific finite element analysis is expanded for pediatric applications.

TBS is a texture parameter related to the fractional Brownian motion (fBm) approach [133]. It is an estimator of the generalized Hurst exponent (Hq) which characterizes what type of process the fBm is. Hq is related to the global behavior of the spatial data (for instance, from the pixel distribution contained in a X-ray image) [134]. Several techniques exist to evaluate Hq [134, 135]. Among them, the variogram is one of the most popular approaches based on variance for the evaluation of the average trend [136]. Hq is calculated at the slope of an interpolated straight line in the log–log system of the variogram [136, 137]. TBS uses a custom version of the variogram approach (Fig. 13.5). Although TBS seems to be an estimator of Hq [136, 137], it is not one, owing to some “black box” differences (patented method and industrial secrets).

A part of TBS algorithm has been originally described by Pothuaud et al. [130] and subsequently enhanced to obtain the actual version of the algorithm [131, 132]. TBS has been designed to analyze images provided by DXA devices. It is evaluated directly from the raw data of the device sensor (i.e. DXA sensor). It characterizes the rate of variation in the gray levels of the 2D projection image and is expressed without units (see Fig. 13.6). Basically, it takes into account gray level amplitudes, the number of these amplitudes and their distributions over the DXA image.

DXA image texture is linked to the texture of the projected bone but also to the acquisition “noise” which is mainly due to the soft tissues above and under the bone in the region of interest. This noise negatively impacts TBS [132], that is, the thicker the soft tissue, the greater the “noise,” which lessens TBS. To overcome this effect, a soft tissue correction based on subject’s body mass index (BMI), has been implemented and validated for adults.

TBS is calculated from PA lumbar spine DXA images only. Using such imaging techniques, it is not possible to directly measure classical 3D bone microarchitecture parameters. However, gray level variations—as evaluated on 2D DXA projected images—reflect global variations in X-ray absorption properties in the corresponding 3D tissue microarchitecture. TBS is an indirect evaluation of the 3D structure; i.e. TBS correlates with some 3D bone microarchitecture parameters such as bone volume fraction, trabecular bone number, trabecular separation, connectivity density, and Structure Model Index at vertebrae [131, 132] but also at the radius [138–140] or at iliac crests [141].

TBS can be evaluated using all available DXA images acquired on Hologic (Delphi™, Discovery™, Horizon™ or QDR4500™ series; Bedford, MA, USA) or GE-Lunar DXA devices (Prodigy™ or iDXA™ series; Madison, WI, USA). TBS is calculated using the same region of interest (ROI) as the aBMD measurement: if a vertebra is excluded from the aBMD evaluation (ex: presence of a fracture), the same vertebra is automatically excluded from the TBS analysis. A TBS value is calculated for each bone pixel of the considered ROI. A TBS value is provided by vertebra composing the ROI (average of the pixel TBS values) as well as the TBS value for the entire ROI (average of the vertebrae TBS values, as presented in Fig. 13.6). TBS is reported without unit. TBS is embedded into TBS iNsight™ software (Medimaps SASU, Mérignac, France) a CE marked and FDA 510 k cleared tool used for clinical purposes in adults.

To date, more than 100 articles have been published examining the information gleaned from TBS in adults with osteoporosis [142, 148]. The findings suggest that: (1) TBS is able to predict osteoporotic fracture (osteoporotic hip fractures or major osteoporotic fractures) as well as the aBMD but independently of aBMD and Clinical Risk Factors (CRF) [143–147]; (2) TBS in combination with aBMD improves fracture prediction [143–145]; (3) TBS is an independent CRF [146]; (4) TBS can assess the fracture risk in some causes of secondary osteoporosis [40]: diabetes [149–151], glucocorticoid induced osteoporosis [152, 153] or hyperparathyroidism [154, 155]; (5) TBS is not impacted by lumbar spine osteoarthritis [156, 157]; and (6) TBS exhibits a different response upon treatment types [158–160]. Recommendations generated by several international scientific societies have suggested that TBS be included in clinical practice for adults only [148, 161, 162].

From a technical point of view, TBS has been optimized for use in adults. More particularly, soft tissue effects on TBS have been evaluated and compensated based on BMI of a patient as a surrogate for tissue thickness. This approach is not appropriate for use in growing children because of changes in body composition, bones, or muscles during growth which vary by individual. In addition, these body modifications are gender-dependent and are neither linear nor uniform with aging. It is therefore important to assess TBS without the adult soft tissue compensation, and simply report a raw TBS value given no dedicated soft tissue correction algorithms yet exist for children. Future strategies could apply a correction using theoretical tissue thickness effect on TBS (using ex vivo data for instance) instead of BMI.

Although widely characterized, used and understood in adults, TBS is at its beginnings in children. Sparse data have been published or presented [163–170] and some of these studies have used the TBS software version dedicated for adults [163, 164]. Consequently, results and conclusions obtained in these two studies have to be interpreted with caution and will not be discussed here.

Other studies [165–170] have used either raw TBS data [165–167] or raw TBS data with a tissue correction based on spine tissue thickness [168–170]. Positive relationships between TBS and the Tanner stage have been observed in both girls [165, 170] and boys [144]. Similarly, positive correlations have been observed between TBS and aBMD or estimates of volumetric BMD or the Bone Mineral Content (BMC) [165–170]. These results are consistent with TBS normative data previously presented [167]. However, it seems that the strength of this association is gender-dependent [165, 167] and also age-dependent [167]. Few data exist in infants [166]. Using a cohort of 109 and 143 healthy male and female infants (aged between birth to 2 years old), the authors [166] observed “U shaped” age-related TBS curves (a decrease followed by an increase) which could be explained by the reorganization of the trabecular structure; trabeculae changed from a radial orientation to a vertical/horizontal orientation [171] in response to changes in the mechanical loading of the spine. Vertebrae are altered with development from a bedrest phase (with small amount of mechanical loading) followed by sitting and then standing phases where weight load is applied. These preliminary reference raw data [166, 167] are a first step. However, as with pediatric norms for aBMD, different factors of the growing skeleton influencing the results have to be investigated and compensated for. Especially, it is important to determine if parameters that influence DXA measurements for aBMD also affect the TBS score. For example, aBMD measurements are influenced by height and gender of patients. Tall children tend to have a higher aBMD compared to small children regardless of their volumetric (“true”) bone mineral density. Reference values for aBMD are adjusted for the hormonal and pubertal status of the individual. In male adolescents an increase of testosterone during puberty leads to an increase of muscle mass which is the most important predictor of bone mass in boys. In pubertal girls the increased estrogen levels inhibit remodeling processes on the endosteal surface resulting in an increased bone mass endosteal compared to boys. These gender-dependent changes during puberty may influence the TBS results and have to be considered. Another difference between children and adults is the presence of growth plates in children. There is very limited knowledge about the influence of the increased amount of cartilage on DXA measurements and this has to be reflected in the interpretation of TBS measurements in children and adolescents.