PDF )   Rev Osteoporos Metab Miner. 2021; 13 (Supl 2): S11-17

Del Río Barquero L
CETIR Medical Center. Ascires Group (Spain)


The incorporation of dual-energy radiological absorptiometry (DXA) to the arsenal of diagnostic techniques unleashed a cascade of advances in the management of metabolic osteopathies. The extensive use of DXA has made it possible to recognize its indications and detect limitations in the evaluation of the risk of bone fracture. In the last decade new advances have been developed applicable to the original technique. These are the Trabecular Bone Score (TBS) and the 3D reconstruction of the DXA images. With different approaches, they allow to assess the microarchitecture of the trabecular bone (TBS) and the measurement with great accuracy of the trabecular and cortical bone, reaching the measurements of volumetric bone density, cortical thickness and geometric variables. This new information now accessible allows the calculation of subject-specific bone strength and opens the possibility of predicting biomechanical behavior in the face of trauma and overload, advancing the diagnosis of fragility before the appearance of fractures.


In 1994, the WHO defined criteria for the diagnosis of osteoporosis using the measurement of bone mineral density (BMD). The DXA technique has established itself as the dominant technology for quantifying BMD due to:
a) strong correlation between BMD measured by DXA and bone strength in biomechanical studies,
b) Epidemiological studies that show a strong relationship between the risk of fracture and BMD,
c) For its use in clinical trials of treatments for the selection of subjects and monitoring based on its excellent precision and low radiation dose.
DXA is indicated to diagnose osteoporosis, assess fracture risk, and monitor changes in BMD over time. In recent years, there have been improvements to the initial DXA technology and it is used for other measurements beyond BMD (eg, femur geometry, vertebral fracture detection, body composition analysis).
One of the limitations in the diagnostic approach to bone mass measurement is the overlap of BMD values in subjects with and without fractures[1]. The predictive value of the BMD measurement is limited and the WHO diagnostic threshold of the T-score <-2.5 is taken as one more factor (albeit the most powerful), but not the only one, in taking of clinical decisions. The last definition of osteoporosis included the concept of alteration of the structural quality of bone.
Bone quality encompasses multiple factors not directly related to bone mass. Macro and microstructural factors have been identified due to their relationship with bone strength and therefore with fragility fractures. To satisfy the need for evaluation of these structural factors, new bone evaluation procedures have been developed using the most widely used technique, DXA. This document will briefly review the application of the TBS technique (Trabecular Bone Score, index or score of the trabecular bone) and the 3D reconstruction of the DXA image, which opens up new horizons such as the calculation of bone strength with a non-standard method simple and safe invasive.

TBS physical fundamentals
The TBS has been described by the developers as an image texture parameter that reflects the differences in the level of gray with which the cells (pixels) are represented in DXA images. The TBS is calculated using the raw data from the DXA acquisition but the calculation is done separately and by different methods than the BMD. The TBS is processed upon completion of the measurement and analysis of the BMD scan and applied to the same region of interest. The TBS calculation principle was published in 2008[2]. The variations in the gray scale with which the contiguous pixels are represented, in multiple random directions constitute the experimental variogram. A 3D image of a narrow network of trabeculae produces a 2D projection image with many small amplitude gray level variations and thus a steep slope of the variogram which offers a high TBS value (a conserved microarchitecture associated with good mechanical resistance). In contrast, a low TBS value indicates low-quality microarchitecture with few gray-level variations, of considerable amplitude, inherent in a gentle slope at the origin of the variogram. The TBS calculation software (TBS InSight[®]) is an application distributed by Med-imaps (France) (Figure 1).
Correlation between TBS and bone microarchitecture parameters:
The correlations in the initial study[2] between TBS and the main 3D micro-architectural parameters measured in trabecular bone samples using micro-CT in different skeletal bones. The following microarchitecture parameters were determined: bone volume/total volume (BV/TV), trabecular thickness (TbTh), trabecular spacing (TbSp), number of trabeculae (TBN), and their connectivity (Conn.) (Table 1)

TBS in aging:
The first TBS development curves in relation to age were established on the results of 5,942 French women[3]. At present, normative studies have been carried out in several countries, including in Spain (SEIOMM-TBS Project) confirming a great similarity in TBS values in both sexes[4-7]. In the Spanish population (SEIOMM-TBS Project), TBS values in adult women and men in the 20-30 year age bracket were very similar and reached their highest value. TBS decreases with age in both sexes. The decline in TBS and BMD is similar in the 40-50 years. In women the average decrease in TBS between 20-80 years was -18% and in men it was -14%. TBS values showed poor correlation with body mass index (r=-0.1), weight (r=-0.1) and L1-L4 BMD (r=0.2) (Figure 2).

The coefficient of variation is similar to the BMD measurements for DXA, being 1.5% for TBS (1.2% for BMD)[8]. The in vivo reproducibility of TBS using the ISCD protocol in 30 unselected patients (26 women and four men) who presented no detectable vertebral fractures. The mean TBS was 1.239±0.082, the coefficient of variation was 1.9%, and the least significant difference was 0.065. For the same patients, the coefficient of variation was 1.2% for BMD.

Studies on the fracture-discrimination capacity of TBS:
Several studies have evaluated the ability of TBS to differentiate patients with fragility fractures from those without fractures. The TBS in all of them was significantly lower in patients with fractures than in controls. These cross-sectional studies indicate that TBS can discriminate individuals with fractures from controls. This discriminatory power of TBS is similar or greater than that of BMD and that the combination of TBS and BMD provide better discrimination than that of BMD. Solo BMD[8-11].
The most cited population study with TBS is the one carried out in the Manitoba Cohort. Lumbar spine TBS and BMD were prospectively compared in a large female population of 29,407 women older than 50 years[12]. For a given lumbar BMD range (normal or osteopenia or osteoporosis), the annual number of incident fractures was always higher in the lowest tertile TBS. Lumbar spine BMD and TBS were weakly correlated (r=0.32). The results were similar for the prediction of hip fractures or any of the four main types of fracture considered. For the four types of fracture together, the predictive ability improved significantly when BMD and TBS were combined.
In the OFELY[13] and OPUS[14] cohort studies, TBS performance was significantly better than lumbar spine BMD for predicting clinical osteoporotic fractures. In radiological vertebral fractures, TBS and BMD of the spine had a similar predictive power. The combination of TBS and BMD increased performance, however, with a predictive capacity similar to the BMD of the total area of the femur and femoral neck. In non-osteoporotic women, TBS predicted incident fragility fractures similarly to BMD. The combination of TBS and BMD improved prediction in all scenarios compared to the use of BMD alone.

Impact of Osteoarthritis on TBS:
Osteoarthritis-related bone sclerosis generates more or less marked contrasts with adjacent healthy bone, and the method used to calculate TBS detects these interfaces and appears only minimally affected by large masses of osteoarthritic bone. In a retrospective cross-sectional study of 141 densitometries in patients with osteoarthritis, only in L4 (using the ISCD criterion, that is, a difference of more than 1 SD between the BMD values between L4 and L3) and 249 controls (defined using ISCD criteria such as BMD of L1 <L2 <L3, and> L4), showed that osteoarthritis had no significant effect on TBS values, as long as the resulting increase in BMD was less than 3.5 standard deviations in L3[3].

TBS as a new risk factor for FRAX[®]:
The developers of TBS and the University of Sheffield research group determined the impact of TBS on the probability of fracture, beyond that provided by the clinical risk factors used in the FRAX tool. The Manitoba, Canada cohort[15] was used in a retrospective study applying the TBS to the initial DXA scan and the rest of the risk variables already used in the FRAX. When fully adjusted for FRAX risk variables, TBS remained a statistically significant predictor of major osteoporotic fractures. Fit models have been derived for the main fractures and hip fractures, taking into account TBS and age. TBS has been found to be a predictor of the risk of osteoporotic fracture, independent of BMD of the femoral neck and clinical risk factors, being a risk factor for mortality.

Application in clinical practice:
The assessment of bone microarchitecture allows the identification of patients at high risk of fracture who have not been adequately classified only by bone mineral density. The application of TBS facilitates the management of patients by recognizing subjects with low BMD and altered bone structure[16]. In this sense, the population sector that can benefit most from the application of the TBS are those who have a BMD T-score <-1.0 and >-2.5 (Figure 3).
In the routine clinical practice, TBS should be considered as an additional “risk factor” that will help in the orientation and management of the patient at risk of osteoporosis. Since TBS is less affected by the artifacts that most influence BMD measurements, whether intrinsic, such as degenerative alterations, extra-skeletal calcifications, or extrinsic, such as orthopedic elements, they increase the diagnostic performance of the DXA technique.

Diagnostic thresholds:
The reference values used have been re-validated after the review of several cohorts of the European female population, which also included women from Spain. With the current data, a threshold of significant degradation of the microarchitecture of the trabecular bone is considered, a TBS result lower than 1,200.

The TBS is a new method of application of the DXA diagnostic technique that allows the evaluation of bone microarchitecture, a key determinant of bone strength. The TBS can be calculated in a simple way, using the widely available DXA technology and following the same conventional procedure for the measurement of BMD. The TBS is a reproducible quantitative value and therefore can be monitored. The clinical results obtained in large population groups confirm that the combination of BMD and TBS is capable of predicting fragility fractures and thus substantially improves the predictive capacity of fracture risk.

Although DXA accurately measures BMD, it is limited by its two-dimensionality and does not represent the spatial distribution of BMD in the bone structures examined. To overcome this limitation, quantitative computed tomography (QCT) allows 3D reconstruction and assessment of the distribution of BMD in bone. Several parameters evaluated in 3D are strongly correlated with the strength of the femur, such as trabecular and cortical BMD, volumetric BMD in specific regions, or geometric parameters such as the length of the neck axis and the cortical thickness, and justify the variation in force. necessary to achieve mechanical failure, improving the estimation of the risk of fracture when it complements the BMD[17]. However, QCT exposes a significantly higher radiation dose than DXA and has a higher cost, reasons that limit its application to research areas.
In recent years, it has been possible to reconstruct a statistical model that combines 3D bone shape and BMD distribution from an in vivo database of QCT scans of the proximal femur[18,19]. Reconstruction is performed using an intensity-based 3D-2D registration process and the similarity between your QCT projection and the DXA image is maximized. The methodology places emphasis on achieving a 3D reconstruction of the DXA image with a bone density model that resembles, through iterative updating, the information that a QCT would have, resolving the relationship with other bone structures (Figure 4).
This method has been specified in commercial software known as 3D-Shaper[® ](Galgo Medical, Barcelona, Spain), available for application in most of the currently existing densitometers (GE-Lunar, Hologic, DMS). One of the most interesting aspects of this technique is that it uses the DXA image obtained in a conventional way, without conditioning a different procedure or longer or higher radiation dose. It is currently available for 3D reconstruction of the proximal femur and lumbar spine.

Variables that it calculates:
The software allows 3D reconstruction and volumetric bone density (cm[3]) measurements in isolation from the trabecular component, from the cortical bone or by integrating both compartments (integrated bone): trabecular volumetric BMD (trabecular vBMD, in mg/cm[3]), volumetric BMD cortical (cortical vBMD, in mg/cm[3]), volumetric bone integrated BMD (global vBMD, in mg/cm[3]). The measurement of cortical thickness (Cth) in mm and the cortical surface BMD (cortical sBMD) is calculated, at each vertex of the femoral surface mesh, as the multiplication of the Cth (in cm) by the cortical vBMD ( in g/cm[3]) observed throughout this thickness. Cortical sBMD is expressed in grams per square centimeter. Cortical sBMD has been used in studies using QCT in the literature[20]. In the context of patient monitoring, any increase in Cth, cortical vBMD, or both will result in an increase in cortical sBMD. On the other hand, if Cth and cortical vBMD vary similarly in opposite ways (eg, increased Cth and decreased cortical vBMD), the cortical sBMD will remain unchanged.
If the region in question is the femur, the mean values of the aforementioned variables are calculated on the total region of interest of the femur (neck, trochanter, inter-trochanteric region, diaphysis and total area), or in the lumbar spine (in the vertebrae L1 to L4 and their averages)[21].

The accuracy of the measurements and the potential need for a single or multiple bone region scan (different angles) was evaluated in vivo by comparing 3D reconstructions obtained from simulated DXA images using repeated DXA scans with patient repositioning and different inclinations, with 3D QCT reconstructions. The comparison showed that the use of a single DXA provides highly accurate 3D reconstructions (mean shape precision of 1.0 mm and BMD distribution errors of 7.0%)[22]. In another study, high-resolution micro-CT data from 23 proximal cadaver femurs were analyzed to determine a relationship between cortical thickness and density[23] and was complemented with a case-control study that included patients with osteoporosis. and age-matched controls with normal bone density to evaluate the method in a clinical setting. Cortical thickness (density) estimation errors were 0.07±0.19 mm (-18±92 mg/cm[3]) using simulated clinical CT volumes with the smallest voxel size (0.33×0,33×0.5 mm[3]), and 0.10±0.24 mm (-10±115 mg/cm[3]) using the volumes with the largest voxel size (1.0×1.0×3.0 mm[3]). The case-control study showed that osteoporotic patients had a thinner cortex and lower cortical density, with mean differences of -0.8 mm and -58.6 mg/cm[3] in the proximal femur compared to controls of the same age (p value <0.001).
In the lumbar spine, the accuracy error of the anatomical shape was 1.5 mm in the total vertebra and 0.6 mm in the vertebral body. The correlation coefficients between the measurements derived from DXA and QCT ranged from 0.8 to 0.9[21].

The short-term in vivo precision of 3D measurements made with DXA acquisitions made with HOLOGIC and GE systems in patient follow-up at an 18-month interval between baseline examination and monitoring DXA[24]. Considering the minimum significant change for a 95% confidence interval, the recommended time intervals for the evaluation of trend in postmenopausal women were 2.9 years (integral volumetric BMD), 2.6 years (trabecular volumetric BMD) and 3.5 years (cortical surface BMD), using the Lunar iDXA densitometer. The trend assessment intervals for area BMD were 2.8 years in the neck and 2.7 years in the total femur. The time intervals in postmenopausal women were similar to those measured for 2D αMD measurements in the femur region.

Relationship of 3D parameters with bone strength
In vitro studies:
The method evaluated, in an experiment with bone samples ex vivo, the predictive capacity of bone strength in biomechanical examinations of 90 femurs from cadavers that were previously explored with DXA, obtaining a correlation coefficient of 0.85 between the load of predicted and measured fracture, while a regression using BMDa (DXA) measurements resulted in a correlation coefficient of 0.77[25].

In vivo studies:
In a retrospective case-control study[26], 3D-DXA measurements were evaluated in a cohort of postmenopausal women with hip fracture. The total BMD of the total hip area of the fracture group was 10% lower compared to the control group. The differences in volumetric BMD (total hip) measured by 3D-Shaper were more pronounced in the trabecular compartment (-23%) than in the cortical compartment (-8%). The area under the curve (ROC curves) was 0.742 for trabecular volumetric BMD, 0.706 for cortical volumetric BMD, and 0.712 for total hip area BMD. Differences in the cortex were locally more pronounced on the median aspect of the shaft, the lateral aspect of the greater trochanter and the supero-lateral aspect of the neck. Marked differences in volumetric BMD were observed in the greater trochanter (Figure 5).
In a case study (61 hip fractures) and controls, the predictive capacity of fractures was assessed from 3D measurements of the lumbar spine, evaluating the association of 3D-DXA measurements of the lumbar spine in subjects who had suffered osteoporotic femur fractures. A stronger association was found between femoral neck fractures and lumbar spine cortical bone variables compared to trabecular bone measurements[27].
The association of 3D-DXA measurements with vertebral fractures was evaluated in a retrospective case-control study[28]. Lumbar spine DXA scans were acquired at baseline (ie, before the fracture event for fractured subjects). The BMD of the fracture group was 9.3% lower compared to the control group (p<0.01). However, a greater difference was found for trabecular BMD in the vertebral body (-16.1%, p<0.001), better discriminating the fracture and control groups, with an AUC of 0.733, compared to 0.682 for BMD. This study showed the ability of 3D-DXA measurements to discriminate patients with vertebral fractures and patients who have not suffered them (Figure 6).

Future projection:
The 3D reconstruction of bone regions with complex geometry can benefit from biomechanical analyzes based on finite elements (FE) that help improve the prediction of the risk of fracture by integrating definitive information on the biomechanical behavior when the bone is subjected to physical loads. In this sense, it is worth highlighting the study in which patients with a recent femur fracture and controls were included[29]. A lateral fall was simulated using a maximum static load that depended on the patient’s mass and height. The results showed that the main maximum stress biomechanical variable was a better discriminator (AUC >0.80) than the volumetric BMD (AUC ≤0.70). A high discrimination capacity was achieved when the analysis was carried out by bone type, fracture zone and gender/sex (AUC of 0.91 for women, trabecular bone and trochanter area). The results suggested that the trabecular bone is essential for the discrimination of femur fractures. The application of finite element analysis to models derived from DXA scans can significantly improve the prediction of the risk of fracture of complex sectors such as the femur; providing a new perspective for clinicians to use this new technology.

Conflict of interest: The author declares that he has no conflicts of interest.



1. Cummings SR (1985) Are patients with hip fractures more osteoporotic? Review of the evidence. Am J Med. 78:487-494.
2. Pothuaud L, Carceller P, Hans D (2008). Correlations between grey-level variations in 2D projection images (TBS) and 3D microarchitecture: applications in the study of human trabecular bone microarchitecture. Bone. 42:775-787.
3. Dufour R, Winzenrieth R, Heraud A, Hans D, Mehsen N. Generation and validation of a normative, age-specific reference curve for lumbar spine trabecular bone score (TBS) in French women. Osteoporos Int. 2013 Nov;24(11):2837-46.
4. Guagnelli MÁ, Winzenrieth R, Deleze M, Cons-Molina F, Clark P. Description of Normative Spine TBS Data for Men and Women in Mexican Population. J Clin Densitom. 2020 Jul 2:S1094-6950(20) 30094-9.
5. Simonelli C, Leib E, Mossman N, Winzenrieth R, Hans D, McClung M. Creation of an age-adjusted, dual-energy x-ray absorptiometry-derived trabecular bone score curve for the lumbar spine in non-Hispanic US White women. J Clin Densitom. 2014 Apr-Jun;17(2):314-9.
6. Iki M, Tamaki J, Sato Y, Winzenrieth R, Kagamimori S, Kagawa Y, Yoneshima H. Age-related normative values of trabecular bone score (TBS) for Japanese women: the Japanese Population-based Osteoporosis (JPOS) study. Osteoporos Int. 2015 Jan;26(1):245-52.
7. Looker AC, Sarafrazi Isfahani N, Fan B, Shepherd JA. Trabecular bone scores and lumbar spine bone mineral density of US adults: comparison of relationships with demographic and body size variables. Osteoporos Int. 2016 Aug;27 (8):2467-75.
8. Hans DB, Cormier C, Bloch JG, Dufour R, Héraud A, Barthe N, Colson F, Giraldi JM, Lamy O, Krieg MA (2009) Indice TBS: la microarchitecture par DXA. Impact Santé, Abstract Rhumatologie. 302:4-8.
9. Pothuaud L, Barthe N, Krieg MA, Mehsen N, Carceller P, Hans D (2009). Evaluation of the potential use of trabecular bone score to complement bone mineral density in the diagnosis of osteoporosis: a preliminary spine BMD-matched, case-control study. J Clin Densitom. 12:170-176.
10. Rabier B, Héraud A, Grand-Lenoir C, Winzenrieth R, Hans D (2010). A multicentre, retrospective case-control study assessing the role of trabecular bone score (TBS) in menopausal Caucasian women with low areal bone mineral density (BMDa): analysing the odds of vertebral fracture. Bone. 46:176-181.
11. Winzenrieth R, Dufour R, Pothuaud L, Hans D (2010). A retrospective case-control study assessing the role of trabecular bone score in postmenopausal Caucasian women with osteopenia: analyzing the odds of vertebral fracture. Calcif Tissue Int. 86:104-109.
12. Hans DB, Goertzen AL, Krieg MA, Leslie WD (2009). Bone micro-architecture assessed by TBS predicts clinical spine fractures independently of BMD in 29407 women aged 50 and older: the Manitoba Prospective Study. ASBMR, Denver.
13. Boutroy S, Hans DB, Sornay-Rendu E, Vilayphiou N, Winzenrieth R, Munoz F, Chapurlat R (2010). Trabecular bone score helps classifying women at risk of fracture: a retrospective analysis of the OFELY study. ASBMR, Toronto.
14. Briot K, Paternotte S, Kolta S, Eastell R, Reid DM, Felsenberg D, Glüer CC, Christian Roux C. Added value of trabecular bone score to bone mineral density for prediction of osteoporotic fractures in postmenopausal women: The OPUS study. Bone. 2013 Nov; 57(1):232-6.
15. McCloskey EV, Oden A, Harvey NC, Leslie WD, Hans D, Johansson H, Kanis JA. Adjusting Fracture Probability by Trabecular Bone Score. Calcif Tissue Int. DOI 10.1007/s00223-015-9980.
16. https://seiomm.org/wp-content/uploads/2020/04/Posici%C3%B3n-Oficial-SEIOMM-sobre-TBS.pdf.
17. V. D. Bousson, J. Adams, K. Engelke, M. Aout, M. Cohen-Solal, C. Bergot, D. Haguenauer, D. Goldberg, K. Champion, R. Aksouh, E. Vicaut, and J.-D. Laredo, “In vivo discrimination of hip fracture with quantitative computed tomography: Results from the prospective European Femur Fracture Study (EFFECT),” J. Bone Mineral Res. vol. 26, no. 4, pp. 881-893, 2011.
18. T. Whitmarsh, L. Humbert, M. S.D. Craene, L.M. del Río Barquero, K. Fritscher, R. Schubert, F. Eckstein, T. M. Link, and A. F. Frangi. 3D bone mineral density distribution and shape reconstruction of the proximal femur from a single simulated DXA image: An in vitro study, in Proc. SPIE Med Imag. 2010: Image Process. 2010, vol. 7623, no. 1, p. 76234U.
19. Whitmarsh T, Humbert L, De Craene M, Del Rio Barquero LM, Frangi AF. Reconstructing the 3D shape and bone mineral density distribution of the proximal femur from dual-energy X-ray absorptiometry. IEEE Trans Med Imaging. 2011 Dec;30(12):2101-14.
20. Nicks KM, Amin S, Melton LJ, et al (2013) Three-dimensional structural analysis of the proximal femur in an age-stratified sample of women. Bone. 55:179-188.
21. Lopez Picazo M, Magallon Baro A, Del Rio Barquero LM, Di Gregorio S, Martelli Y, Romera J, Steghofer M, Gonzalez Ballester MA, Humbert L. 3-D Subject-specific shape and density estimation of the lumbar spine from a single anteroposterior DXA image including assessment of cortical and trabecular bone. IEEE Trans Med Imaging. 2018 Dec;37(12):2651-2662.
22. Humbert L, Whitmarsh T, De Craene M, del Río LM,. Frangi A. Technical Note: Comparison between single and multiview simulated DXA configurations for recons- tructing the 3D shape and bone mineral density distribution of the proximal femur. Medical Physics. 39,5272(2012).
23. Humbert L, Hazrati Marangalou J, Del Río Barquero LM, van Lenthe GH, van Rietbergen B. Technical Note: Cortical thickness and density estimation from clinical CT using a prior thickness-density relationship. Med Phys. 2016 Apr;43(4):1945.
24. Humbert L, Winzenrieth R, Di Gregorio S, Thomas T, Vico L, Malouf J, Del Río Barquero LM. 3D Analysis of cortical and trabecular bone from hip DXA: precision and trend assessment interval in postmenopausal women. J Clin Densitom. 2019 Apr-Jun;22(2):214-218. doi: 10. 1016/j.jocd.2018.05.001. Epub 2018 May 8. PMID: 30017573.
25. Humbert L, Whitmarsh T, Fritscher K, del Río LM, Eckstein F, Link T, Schubert R, Frangi A. Femoral strength prediction using a 3d reconstruction method from dual-energy x-ray absorptiometry. 2012 9th IEEE International Symposium on Biomedical Imaging (ISBI).
26. Humbert L, Bagué A, Di Gregorio S, Winzenrieth R, Sevillano X, González Ballester MÁ, Del Rio L. DXA-Based 3D analysis of the cortical and trabecular bone of hip fracture postmenopausal women: a case-control study. J Clin Densitom. 2020 Jul-Sep;23(3):403-410.
27. López Picazo M, Humbert L, Winzenrieth R, Di Gregorio S, González Ballester MA, Del Río Barquero LM. Association between osteoporotic femoral neck fractures and DXA-derived 3D measurements at lumbar spine: a case-control study. Arch Osteoporos. 2020 Jan 3;15(1):8.
28. López Picazo M, Humbert L, Di Gregorio S, González Ballester MA, Del Río Barquero LM. Discrimination of osteoporosis-related vertebral fractures by DXA-derived 3D measurements: a retrospective case-control study. Osteoporos Int. 2019 May;30(5):1099-1110.
29. Ruiz Wills C, Olivares AL, Tassani S, Ceresa M, Zimmer V, González Ballester MA, Del Río LM, Humbert L, Noailly J. 3D patient-specific finite element models of the proximal femur based on DXA towards the classification of fracture and non-fracture cases. Bone. 2019 Apr; 121:89-99.