Commercially available QUS devices for osteoporosis diagnosis are

Commercially available QUS devices for GSK2656157 diagnosis are currently applicable only on peripheral bone regions, and their recognized value is actually limited to fragility fracture prediction in patients >65 y through calcaneal measurements, whose outcome has to be employed in conjunction with clinical risk factors (ISCD 2013). One of the main limitations of peripheral QUS systems is their poor or moderate correlation with spinal DXA outputs (Dane et al. 2008; El Maghraoui et al. 2009; Gemalmaz et al. 2007; Iida et al. 2010; Kwok et al. 2012). However, as expected, the adoption of an innovative US approach exploiting site-matched vertebral measurements resulted in markedly improved correlation. In fact, by reviewing literature quantifying the correlation of current peripheral QUS techniques with spinal DXA results through clinical studies in cohorts of women, we found that r2 was always <0.38 (apart from the very recent article by Jiang et al. [2014], who adopted an experimental backscatter technique for calcaneal measurements and obtained r2 = 0.56 with DXA-measured spinal BMD), emphasizing the value of our reported results. We also documented a measurement precision (RMS-CV = 2.95%) that is comparable to the typical values reported for clinically available peripheral QUS devices (Njeh et al. 2000), but is coupled with the aforementioned higher accuracies. It is important to note that the US signal portions used in this study for spectral model constructions and O.S. value calculations are essentially related to the trabecular part of vertebrae: The development of an extended data analysis protocol, capable of taking into account cortical properties as well, could provide even better correlations with DXA-measured BMD, as preliminarily illustrated by a very recent pilot study focused on “ex vivo” QUS assessment of femoral strength (Grimal et al. 2013). Furthermore, one should consider that even if fracture discrimination was not explicitly involved in the present study, given the significant correlations obtained between O.S.-derived BMD values and the corresponding data provided by DXA, currently representing the gold standard technique for the estimation of bone fragility and fracture risk, our approach can reasonably be expected to perform similarly to DXA in fracture risk assessment as well. Moreover, it is worth observing that the highly selective automatic identification of vertebrae and related ROIs, combined with the significant statistical basis of our proposed approach (requiring the described series of averaging and normalization operations on signals and spectra), has the potential to at least partially overcome random interference noise, one factor limiting the precision of US backscatter measurements. Image and signal selections and the sequences of averaging operations reduce the incidence of any kind of random effect and are also an indirect way to take into account, as a first approximation, that US velocity can vary between different vertebrae and different patients. Actually, our proposed approach, in its present implementation, differs from previously reported approaches because it is based on overall correlations between different spectra, each considered as a whole without extracting any synthetic parameter and without associating a specific meaning to single spectrum peaks or valleys. All the local characteristics of the considered spectra are indirectly taken into GSK2656157 account by the illustrated correlation process, intrinsically providing a more statistically significant basis of the reference data analysis. This, coupled with the statistical derivation of reference models starting from real human data, is probably the reason for the improvement in the correlation between DXA-measured BMD and O.S.-based estimates with respect to different US parameters reported in the literature (e.g., spectral centroid shift), involving only specific spectral features whose values are typically compared with phantom measurements.