Optimal Prediction of Bone Mineral Density with Ultrasonic Measurements in Excised Human Femur

Bone mineral density (BMD) measured with dual energy X-ray absorptiometry (DXA) techniques is the current gold standard for osteoporotic fracture risk prediction. Quantitative ultrasound (QUS) techniques in transmission measurements are, however, increasingly recognized as an alternative approach. It is feasible to select different QUS methods, one type being optimized to assess microarchitectural properties of bone structure and another to assess BMD. Broadband ultrasonic attenuation (BUA) and ultrasonic velocity (UV) measured on the proximal human femur have been shown to be both significantly correlated with BMD. However, a great diversity of algorithms has been reported to measure the time-of-flight used to derive UV values. The purpose of this study was to determine which procedure results in the optimal BMD prediction at the proximal femur from ultrasound measurements. Thirty-eight excised human femurs were measured in transmission with a pair of focused 0.5−MHz central frequency transducers. Two-dimensional scans were performed and radiofrequency (RF) signals were recorded digitally at each scan position. BUA was estimated and eight different signal processing techniques were performed to estimate UV. For each signal-processing technique UV was compared to BMD. We show that the best prediction of BMD was obtained with signal-processing techniques taking into account only the first part of the transmitted signal (r²_BMD-SOS = 0.86). Moreover, we show that a linear multiple regression using both BUA and speed of sound (SOS) and applied to site-matched regions of interest improved the accuracy of BMD predictions (r²_BMD-SOS/BUA = 0.95). Our results demonstrate that selecting specific signal-processing methods for QUS variables allows optimal assessment of BMD. Correlation is sufficiently high that this specific QUS method can be considered as a good surrogate of BMD.