O. Ahmad, K. Ramamurthi, K. Wilson, K. Engelke, R. Prince, R. Taylor

3D assessment of proximal femur geometry and density using volumetric DXA (VXA) may improve assessment of skeletal fragility. In this study we tested the correlation between VXA and 3D QCT derived structural parameters and volumetric density at the proximal femur.
Forty-three women, 82 ± 2.4 years with a BMI less than 35 (a subset of the CARE study) were measured at the proximal femur with a 64 slice Phillips CT scanner (1 mm slice thickness, 0.29 mm in plane resolution, total effective dose 4 mSv) with a Mindways phantom in the field of view. Additionally, four DXA projections of the proximal femur (total effective dose .047 mSv) were taken at -21°, 0°, 20°, and 30°, where the angle is measured from the PA view. These four DXA projections were used to reconstruct a model of the patient’s proximal femur by utilizing information from a statistical model of proximal femurs. The statistical model of proximal femurs was created from QCT’s (1 mm slice thickness, 0.29 mm in plane resolution, GE Lightspeed) of 73 Caucasian women age >65 using methods previously described (Ahmad et al, Abstract W 224, ASBMR 2007). Briefly, an iterative process is performed wherein the DXA projections are compared to projections of the statistical model and the modes of variation of the statistical model are varied to minimize the error between these simulated DXA projections of the model and the actual DXA images acquired. Each resulting patient model (termed a VXA) was then voxelized. We examined the association between the patient VXA and QCT for femoral neck axis length (FNAL), the cross sectional area of the narrowest part of the femoral neck (CSA) and the volumetric density of a 10 mm slice at the narrowest part of the femoral neck (vBMD).
There was a highly significant linear correlation between VXA and QCT for all parameters: FNL (r=0.87, SEE = 0.25 cm, intercept not statistically different from zero, slope not statistically different from one), CSA (r=0.86 SEE = 0.63 cm², intercept not statistically different from zero, slope 1.1 ± 0.02) and vBMD (r=0.80, slope and intercepts different from one and zero due to calibration differences between QCT and DXA).
VXA reconstructions from four DXA views can be obtained readily in vivo with limited radiation exposure. These VXA reconstructions are correlated with geometric parameters and volumetric density measured using 3D QCT.

Disclosures: K. Ramamurthi, Hologic, Inc: Employment (full or part-time). K. Wilson, Hologic, Inc.: Employment (full or part-time). R. Prince, Hologic, Inc.: Research Grants. R. Taylor, Hologic, Inc.: Research Grants. O. Ahmad, Hologic, Inc: Research Grants. K. Engelke, Hologic, Inc.: Research Grants.