Evaluating accuracy of structural geometry by DXA methods with an anthropometric proximal femur phantom

DXA-derived bone structural geometry has been reported extensively but lacks an accuracy standard. In this study, we describe a novel anthropometric structural geometry phantom that simulates the proximal femur for use in assessing accuracy of geometry measurements by DXA or other X-ray methods. The phantom consists of seven different interchangeable neck modules with geometries that span the range of dimensions in an adult human proximal femur, including those representing osteoporosis. Ten repeated hip scans of each neck module using two current DXA scanner models were performed without repositioning. After scanner specific calibration, hip structure analysis was used to derive structural geometry. Scanner performance was similar for the two manufacturers. DXA-derived HSA geometric measurements were highly correlated with values derived directly from phantom geometry and position; R² between DXA and phantom measures were greater than 94% for all parameters, while precision error ranged between 0.3 and 3.9%. Despite high R² there were some systematic geometry errors for both scanners that were small for outer diameter, but increasing with complexity of geometrical parameter; e.g. buckling ratio. In summary, the anthropometric phantom and its fabrication concept were shown to be appropriate for evaluating proximal femoral structural geometry in two different DXA systems.